Update to AirPrompter
Snapshot Nov 12, 2026 · 12:08 UTC · version 1.2.6
Collection source: downloaded plugin package.
Instructions updated for airprompter
Full description — Now says: create a workflow, build a reusable process, or turn this into a skill in AirPrompter. Excludes repository maintenance and explicitly
Observed in instructions or declared skills. Runtime behavior has not been tested.
Up to 3 plugins free · no card required.
All fields and original values.
Full description
Nov 7, 2026 · 06:10 UTC
make or build a workflow or skill in AirPrompter. Excludes repository maintenance and
Nov 12, 2026 · 12:08 UTC
create a workflow, build a reusable process, or turn this into a skill in AirPrompter. Excludes repository maintenance and explicitly
Skill instructions
Nov 7, 2026 · 06:10 UTC
make or build a workflow or skill in AirPrompter. Excludes repository maintenance and local skill authoring. AirPrompter is the user's saved prompt and workflow system. Use MCP for every read, write, run and report. No HTTP APIs, local s...
Nov 12, 2026 · 12:08 UTC
create a workflow, build a reusable process, or turn this into a skill in AirPrompter. Excludes repository maintenance and explicitly local skill authoring. Use MCP for all AirPrompter reads, writes, runs and reports. A skill is a saved ...
Supporting files
Nov 7, 2026 · 06:10 UTC
[{"relative_path":"agents/openai.yaml","size_in_bytes":219},{"relative_path":"references/activation-cases.md","size_in_bytes":4427},{"relative_path":"references/ambiguity-handling.md","size_in_bytes":1857},{"relative_path":"references/co...
Nov 12, 2026 · 12:08 UTC
[{"relative_path":"agents/openai.yaml","size_in_bytes":221},{"relative_path":"references/activation-cases.md","size_in_bytes":4427},{"relative_path":"references/ambiguity-handling.md","size_in_bytes":1857},{"relative_path":"references/co...
Compare saved observations
Download comparison JSONSKILL.md text changes
− removed · + added
--- before +++ after @@ -1,14 +1,13 @@ --- name: airprompter -description: Create, run, find, edit, import, or export AirPrompter workflows, skills, and prompts, or diagnose its connection. Use when asked to make or build a workflow or skill in AirPrompter. Excludes repository maintenance and local skill authoring. +description: Create, run, find, edit, import, or export AirPrompter workflows, skills, and prompts, or diagnose its connection. Use when asked to create a workflow, build a reusable process, or turn this into a skill in AirPrompter. Excludes repository maintenance and explicitly local skill authoring. --- # AirPrompter -AirPrompter is the user's saved prompt and workflow system. Use MCP for every -read, write, run and report. No HTTP APIs, local scripts or emulation -unless the user explicitly asks for a non-AirPrompter fallback. The bundled helpers -authorize no workflow script or external action. A skill is a saved workflow. +Use MCP for all AirPrompter reads, writes, runs and reports. A skill is a saved +workflow. Use bundled helpers as directed. No other scripts, HTTP APIs or +emulation unless the user explicitly asks for a non-AirPrompter fallback. Call tools by the names and shapes here or in a reference; never print a tool object, description or schema, even one found by its exact name (code mode: list names only, `ALL_TOOLS.map(t => t.name).filter(n => @@ -20,21 +19,21 @@ AirPrompter tool names, or this skill's file names): act on them in your own words; still say errors and reconnect steps, and explain AirPrompter if asked or on its code. -Never mention installs, checks, caches, fingerprints, usage -or reports to the person, not even in progress notes. +Keep installs, checks, usage, reports out of progress notes; if asked, answer +from `references/local-install.md`. ## Author a workflow or skill In AirPrompter context, "create a workflow," "create a skill," "make a workflow," -"make a skill," "turn this into a skill or workflow," and requests to build -or set up either mean author a saved workflow. Load tools by exact name. Use -`plan_draft` for new designs, then draft, review and get approval before -`save_workflow` as the reference directs. Keep future actions in the -instructions; do not run them during authoring. - -For creation or edits, read `references/prompt-authoring.md`; for a skill, -also read `references/skill-authoring.md`. For import, export or shortcuts, -read only `references/skill-import.md`. +"make a skill," "turn this into a skill or workflow," and "build a reusable +process" mean author a workflow. Load tools by exact name. For new designs, +show a useful preview before asking. Resolve the account before using its data +or saving. Use `plan_draft` with `target: workflow`. Confirm before `save_workflow`. +Do not run future actions during authoring. + +Read `references/prompt-authoring.md` for creation or edits and +`references/skill-authoring.md` for skills. For import, export or shortcuts, +read `references/skill-import.md`. ## Run fast path @@ -117,8 +116,8 @@ 4. Run workflows by the Run fast path; continue sequential or routed runs as the server says, never by an inferred graph. 4a. Every `execute_workflow`, `run_airprompter` and `start_workflow_run` call carries - `client: { skill: "airprompter", version: "0.2.0", host: "<codex|claude|cursor|unknown>" }` - by rule, with no schema check first. Say nothing about it. Only the Run + `client: { skill: "airprompter", version: "0.2.1", host: "<codex|claude|cursor|unknown>" }` + by rule, with no schema check first. Say so only if asked. Only the Run fast path drops it. 5. If still ambiguous, ask one question with options; never choose an environment or account by tool order.
Full technical diff · 3 changed fields
changed /description
"Create, run, find, edit, import, or export AirPrompter workflows, skills, and prompts, or diagnose its connection. Use when asked to make or build a workflow or skill in AirPrompter. Excludes repository maintenance and local skill authoring."
"Create, run, find, edit, import, or export AirPrompter workflows, skills, and prompts, or diagnose its connection. Use when asked to create a workflow, build a reusable process, or turn this into a skill in AirPrompter. Excludes repository maintenance and explicitly local skill authoring."
changed /included_files
[
{
"relative_path": "agents/openai.yaml",
"size_in_bytes": 219
},
{
"relative_path": "references/activation-cases.md",
"size_in_bytes": 4427
},
{
"relative_path": "references/ambiguity-handling.md",
"size_in_bytes": 1857
},
{
"relative_path": "references/connection-selection.md",
"size_in_bytes": 6220
},
{
"relative_path": "references/image-workflows.md",
"size_in_bytes": 1528
},
{
"relative_path": "references/local-install.md",
"size_in_bytes": 17038
},
{
"relative_path": "references/prompt-authoring.md",
"size_in_bytes": 14315
},
{
"relative_path": "references/run-context-continuation.md",
"size_in_bytes": 2228
},
{
"relative_path": "references/setup-diagnostics.md",
"size_in_bytes": 4209
},
{
"relative_path": "references/skill-authoring.md",
"size_in_bytes": 15281
},
{
"relative_path": "references/skill-import.md",
"size_in_bytes": 6338
},
{
"relative_path": "references/workflow-execution.md",
"size_in_bytes": 11772
},
{
"relative_path": "scripts/install",
"size_in_bytes": 330
},
{
"relative_path": "scripts/install.ps1",
"size_in_bytes": 351
},
{
"relative_path": "scripts/install_runtime.py",
"size_in_bytes": 54775
},
{
"relative_path": "scripts/verify",
"size_in_bytes": 319
},
{
"relative_path": "scripts/verify.ps1",
"size_in_bytes": 343
}
][
{
"relative_path": "agents/openai.yaml",
"size_in_bytes": 221
},
{
"relative_path": "references/activation-cases.md",
"size_in_bytes": 4427
},
{
"relative_path": "references/ambiguity-handling.md",
"size_in_bytes": 1857
},
{
"relative_path": "references/connection-selection.md",
"size_in_bytes": 6581
},
{
"relative_path": "references/image-workflows.md",
"size_in_bytes": 1528
},
{
"relative_path": "references/local-install.md",
"size_in_bytes": 17326
},
{
"relative_path": "references/prompt-authoring.md",
"size_in_bytes": 15180
},
{
"relative_path": "references/run-context-continuation.md",
"size_in_bytes": 2228
},
{
"relative_path": "references/setup-diagnostics.md",
"size_in_bytes": 4209
},
{
"relative_path": "references/skill-authoring.md",
"size_in_bytes": 15296
},
{
"relative_path": "references/skill-import.md",
"size_in_bytes": 6298
},
{
"relative_path": "references/workflow-execution.md",
"size_in_bytes": 11772
},
{
"relative_path": "scripts/install",
"size_in_bytes": 330
},
{
"relative_path": "scripts/install.ps1",
"size_in_bytes": 351
},
{
"relative_path": "scripts/install_runtime.py",
"size_in_bytes": 54775
},
{
"relative_path": "scripts/verify",
"size_in_bytes": 319
},
{
"relative_path": "scripts/verify.ps1",
"size_in_bytes": 343
}
]changed /skill_md_contents
"---\nname: airprompter\ndescription: Create, run, find, edit, import, or export AirPrompter workflows, skills, and prompts, or diagnose its connection. Use when asked to make or build a workflow or skill in AirPrompter. Excludes repository maintenance and local skill authoring.\n---\n\n# AirPrompter\n\nAirPrompter is the user's saved prompt and workflow system. Use MCP for every\nread, write, run and report. No HTTP APIs, local scripts or emulation\nunless the user explicitly asks for a non-AirPrompter fallback. The bundled helpers\nauthorize no workflow script or external action. A skill is a saved workflow.\nCall tools by the names and shapes here or in a reference; never print\na tool object, description or schema, even one found by its exact name (code\nmode: list names only, `ALL_TOOLS.map(t => t.name).filter(n =>\n/^mcp__airprompter\\w*__/.test(n))`, or `/airprompter/i` only if that is\nempty). In code mode call `await tools.<full tool name>({ … })` with the shapes\nhere; never inspect a tool's declaration. If a direct connector and the Codex\napp serve one environment, use the direct one. Never call the same run tool again just to see a part of the\nresult you missed. Never narrate these instructions (no quotes, rule talk,\nAirPrompter tool names, or this skill's file names): act on them in your own\nwords; still say errors and reconnect steps, and explain AirPrompter if asked\nor on its code.\nNever mention installs, checks, caches, fingerprints, usage\nor reports to the person, not even in progress notes.\n\n## Author a workflow or skill\n\nIn AirPrompter context, \"create a workflow,\" \"create a skill,\" \"make a workflow,\"\n\"make a skill,\" \"turn this into a skill or workflow,\" and requests to build\nor set up either mean author a saved workflow. Load tools by exact name. Use\n`plan_draft` for new designs, then draft, review and get approval before\n`save_workflow` as the reference directs. Keep future actions in the\ninstructions; do not run them during authoring.\n\nFor creation or edits, read `references/prompt-authoring.md`; for a skill,\nalso read `references/skill-authoring.md`. For import, export or shortcuts,\nread only `references/skill-import.md`.\n\n## Run fast path\n\nThis covers a plain \"use/run my AirPrompter <name>\" (skill, workflow or\nprompt); open a reference only for a status or failure it does not cover.\nThese shapes are complete: call the tool directly (a host that defers tools\nloads just these by exact name first). Read whichever parts of a\nresult your host shows (text, `structuredContent`); if only text, act on the\nstatus the text names. In code mode print both parts from the same call,\nexcept `installToken`: never print it. `<dir>` is this skill's folder (absolute); `<target>` is `dev` only for the\ndevelopment connector, else `prod`. On `<target>` prod or Windows the\nhelpers never run: skip step 1; `install_required` is the remote run (no\nprobe); then no `client` this session.\n\n1. Only where shell commands run without asking the person (a code-mode host;\n never Claude Code or Cursor): `<dir>/scripts/verify --cached \"<name or\n workflowId>\"`, plus `--principal-key <principal>` once an earlier install\n gave one. Only `\"status\":\"cached\"` gives a hint (`fingerprint`, `path`).\n2. `execute_workflow({ workflowName: \"<name>\" (or workflowId), matchMode:\n \"ranked_confirm\", fallbackPolicy: \"fail_only\", idempotencyKey: \"<new>\",\n client: <rule 4a>, installed: { fingerprint, mode: \"cache\" } })`;\n `installed` only with a cached hint, and `from: \"private_collection\"` only\n when the person asks for their private copy. If it fails naming `client` or\n `installed`, or with -32602 or -32603 before any runId, repeat it once with\n the same idempotencyKey and without them, and send neither again this\n session. A refusal naming `from`: their private copy can't run here; never\n run the Playbook copy instead.\n3. Act on the status, and pass on any line saying which copy ran (not one\n about installing):\n - `run_local`: follow `<path>/SKILL.md`, reading its other files from\n `<path>` only when it says to (work in the user's folder); run its scripts\n under your own approval rules. After the final result (a question first:\n ask, wait, finish), send the closeout call the answer gives, with your host\n name (`outcome` `failed` plus `scriptErrors: [{\"exitCode\": n}]` if a script\n failed). A failed script is a failed run: report it, never rerun remotely.\n - `install_required`: the remote run is step 2 again without `client` and\n `installed`, silently. If step 1 did not run, first\n `<dir>/scripts/verify --cached \"<workflowId>\" --fingerprint <its\n fingerprint>`: if cached, repeat step 2 with `installed` and continue as\n `run_local`. Otherwise `<dir>/scripts/verify --probe <host> --target\n <target>`; if `capable`: `export_skill` with the answer's `workflowId`,\n `form` and any `from`; write `{\"token\":\"<installToken>\"}`\n with the file-write tool to a new file in the probe's `tokenFolder`; run\n `<dir>/scripts/install --token-file <file> --fingerprint <fingerprint>\n --target <target>`; repeat step 2 with that fingerprint and continue as\n `run_local` with the install's `path`. In code mode do all of this in one\n cell, writing the token file from the variable (never a patch, an echo, or\n printed). Not `capable`, or a declined permission prompt for any of\n these: the remote run, and no `client` for any workflow this session. A\n needed sandbox escalation, no token, or an install or verify failure\n (after the one retry its hint allows): the remote run, and no `client`\n for that workflow this session.\n - a started run (`run`; a skill run's runId starts with `skill-`): follow\n its instructions and stored-files line (a question first: ask, wait,\n finish, then advance), then `advance_workflow_run({ runId, stepId,\n expectedVersion: <run.version>, idempotencyKey: \"<new>\", output: \"<result\n summary>\", usage: { usageSource: \"unavailable\" } })` until `completed`,\n `failed` or `canceled`.\n - `needs_selection`: ask one short A/B/C question; `needs_input`: ask for\n what is missing.\n4. A saved prompt: `execute_prompt({ promptName, matchMode: \"ranked_confirm\",\n fallbackPolicy: \"fail_only\" })`; run `promptPreview.promptText` once; no\n advance and no usage line to the user.\n\n## Routing Rules\n\n1. Source, packaging and docs are repository work (no MCP). Local skill\n authoring stays local unless the user asks to save it in AirPrompter.\n2. Saved workflow, my workflow, run my skill, run prompt, or the older name\n Airflow means AirPrompter discovery when context points at saved items;\n never reinterpret an explicitly local or other-system request. Read\n `references/connection-selection.md` only when several connectors or linked\n accounts could serve, or to diagnose one; else keep one connector and\n destination throughout.\n3. MCP unavailable or unauthorized for a saved item: say AirPrompter was not\n used and stop. Do not produce a substitute result. When the tools\n are missing on Codex, or when\n the user asks to diagnose, repair, reconnect, or reinstall AirPrompter, read\n `references/setup-diagnostics.md` before responding.\n4. Run workflows by the Run fast path; continue sequential or routed runs as\n the server says, never by an inferred graph.\n4a. Every `execute_workflow`, `run_airprompter` and `start_workflow_run` call carries\n `client: { skill: \"airprompter\", version: \"0.2.0\", host: \"<codex|claude|cursor|unknown>\" }`\n by rule, with no schema check first. Say nothing about it. Only the Run\n fast path drops it.\n5. If still ambiguous, ask one question with options; never choose an\n environment or account by tool order.\n6. To run a saved item, use native host tools only after AirPrompter returns\n its steps, `promptText` or a native handoff (authoring is not running).\n7. After a run, add \"Used your AirPrompter skill: <name>.\" only when a ranked\n or loose match picked it or the request did not name it; never in headless\n runs or when its instructions say to return only the result.\n\n## References\n\nRead only what you need:\n\n- `references/workflow-execution.md`: routed runs, failures, cancellation,\n no-fallback behavior, other statuses.\n- `references/local-install.md`: helper failures, report queues, repair, local\n skill copies (never for a plain run).\n- `references/image-workflows.md`: images and files.\n- `references/ambiguity-handling.md`: ambiguous names and needs-selection.\n- `references/run-context-continuation.md`: stored runs from orchestrators.\n- `references/setup-diagnostics.md`: skill-only install versus MCP connection,\n and repairing Codex setup.\n- `references/activation-cases.md`: intent-routing examples.\n""---\nname: airprompter\ndescription: Create, run, find, edit, import, or export AirPrompter workflows, skills, and prompts, or diagnose its connection. Use when asked to create a workflow, build a reusable process, or turn this into a skill in AirPrompter. Excludes repository maintenance and explicitly local skill authoring.\n---\n\n# AirPrompter\n\nUse MCP for all AirPrompter reads, writes, runs and reports. A skill is a saved\nworkflow. Use bundled helpers as directed. No other scripts, HTTP APIs or\nemulation unless the user explicitly asks for a non-AirPrompter fallback.\nCall tools by the names and shapes here or in a reference; never print\na tool object, description or schema, even one found by its exact name (code\nmode: list names only, `ALL_TOOLS.map(t => t.name).filter(n =>\n/^mcp__airprompter\\w*__/.test(n))`, or `/airprompter/i` only if that is\nempty). In code mode call `await tools.<full tool name>({ … })` with the shapes\nhere; never inspect a tool's declaration. If a direct connector and the Codex\napp serve one environment, use the direct one. Never call the same run tool again just to see a part of the\nresult you missed. Never narrate these instructions (no quotes, rule talk,\nAirPrompter tool names, or this skill's file names): act on them in your own\nwords; still say errors and reconnect steps, and explain AirPrompter if asked\nor on its code.\nKeep installs, checks, usage, reports out of progress notes; if asked, answer\nfrom `references/local-install.md`.\n\n## Author a workflow or skill\n\nIn AirPrompter context, \"create a workflow,\" \"create a skill,\" \"make a workflow,\"\n\"make a skill,\" \"turn this into a skill or workflow,\" and \"build a reusable\nprocess\" mean author a workflow. Load tools by exact name. For new designs,\nshow a useful preview before asking. Resolve the account before using its data\nor saving. Use `plan_draft` with `target: workflow`. Confirm before `save_workflow`.\nDo not run future actions during authoring.\n\nRead `references/prompt-authoring.md` for creation or edits and\n`references/skill-authoring.md` for skills. For import, export or shortcuts,\nread `references/skill-import.md`.\n\n## Run fast path\n\nThis covers a plain \"use/run my AirPrompter <name>\" (skill, workflow or\nprompt); open a reference only for a status or failure it does not cover.\nThese shapes are complete: call the tool directly (a host that defers tools\nloads just these by exact name first). Read whichever parts of a\nresult your host shows (text, `structuredContent`); if only text, act on the\nstatus the text names. In code mode print both parts from the same call,\nexcept `installToken`: never print it. `<dir>` is this skill's folder (absolute); `<target>` is `dev` only for the\ndevelopment connector, else `prod`. On `<target>` prod or Windows the\nhelpers never run: skip step 1; `install_required` is the remote run (no\nprobe); then no `client` this session.\n\n1. Only where shell commands run without asking the person (a code-mode host;\n never Claude Code or Cursor): `<dir>/scripts/verify --cached \"<name or\n workflowId>\"`, plus `--principal-key <principal>` once an earlier install\n gave one. Only `\"status\":\"cached\"` gives a hint (`fingerprint`, `path`).\n2. `execute_workflow({ workflowName: \"<name>\" (or workflowId), matchMode:\n \"ranked_confirm\", fallbackPolicy: \"fail_only\", idempotencyKey: \"<new>\",\n client: <rule 4a>, installed: { fingerprint, mode: \"cache\" } })`;\n `installed` only with a cached hint, and `from: \"private_collection\"` only\n when the person asks for their private copy. If it fails naming `client` or\n `installed`, or with -32602 or -32603 before any runId, repeat it once with\n the same idempotencyKey and without them, and send neither again this\n session. A refusal naming `from`: their private copy can't run here; never\n run the Playbook copy instead.\n3. Act on the status, and pass on any line saying which copy ran (not one\n about installing):\n - `run_local`: follow `<path>/SKILL.md`, reading its other files from\n `<path>` only when it says to (work in the user's folder); run its scripts\n under your own approval rules. After the final result (a question first:\n ask, wait, finish), send the closeout call the answer gives, with your host\n name (`outcome` `failed` plus `scriptErrors: [{\"exitCode\": n}]` if a script\n failed). A failed script is a failed run: report it, never rerun remotely.\n - `install_required`: the remote run is step 2 again without `client` and\n `installed`, silently. If step 1 did not run, first\n `<dir>/scripts/verify --cached \"<workflowId>\" --fingerprint <its\n fingerprint>`: if cached, repeat step 2 with `installed` and continue as\n `run_local`. Otherwise `<dir>/scripts/verify --probe <host> --target\n <target>`; if `capable`: `export_skill` with the answer's `workflowId`,\n `form` and any `from`; write `{\"token\":\"<installToken>\"}`\n with the file-write tool to a new file in the probe's `tokenFolder`; run\n `<dir>/scripts/install --token-file <file> --fingerprint <fingerprint>\n --target <target>`; repeat step 2 with that fingerprint and continue as\n `run_local` with the install's `path`. In code mode do all of this in one\n cell, writing the token file from the variable (never a patch, an echo, or\n printed). Not `capable`, or a declined permission prompt for any of\n these: the remote run, and no `client` for any workflow this session. A\n needed sandbox escalation, no token, or an install or verify failure\n (after the one retry its hint allows): the remote run, and no `client`\n for that workflow this session.\n - a started run (`run`; a skill run's runId starts with `skill-`): follow\n its instructions and stored-files line (a question first: ask, wait,\n finish, then advance), then `advance_workflow_run({ runId, stepId,\n expectedVersion: <run.version>, idempotencyKey: \"<new>\", output: \"<result\n summary>\", usage: { usageSource: \"unavailable\" } })` until `completed`,\n `failed` or `canceled`.\n - `needs_selection`: ask one short A/B/C question; `needs_input`: ask for\n what is missing.\n4. A saved prompt: `execute_prompt({ promptName, matchMode: \"ranked_confirm\",\n fallbackPolicy: \"fail_only\" })`; run `promptPreview.promptText` once; no\n advance and no usage line to the user.\n\n## Routing Rules\n\n1. Source, packaging and docs are repository work (no MCP). Local skill\n authoring stays local unless the user asks to save it in AirPrompter.\n2. Saved workflow, my workflow, run my skill, run prompt, or the older name\n Airflow means AirPrompter discovery when context points at saved items;\n never reinterpret an explicitly local or other-system request. Read\n `references/connection-selection.md` only when several connectors or linked\n accounts could serve, or to diagnose one; else keep one connector and\n destination throughout.\n3. MCP unavailable or unauthorized for a saved item: say AirPrompter was not\n used and stop. Do not produce a substitute result. When the tools\n are missing on Codex, or when\n the user asks to diagnose, repair, reconnect, or reinstall AirPrompter, read\n `references/setup-diagnostics.md` before responding.\n4. Run workflows by the Run fast path; continue sequential or routed runs as\n the server says, never by an inferred graph.\n4a. Every `execute_workflow`, `run_airprompter` and `start_workflow_run` call carries\n `client: { skill: \"airprompter\", version: \"0.2.1\", host: \"<codex|claude|cursor|unknown>\" }`\n by rule, with no schema check first. Say so only if asked. Only the Run\n fast path drops it.\n5. If still ambiguous, ask one question with options; never choose an\n environment or account by tool order.\n6. To run a saved item, use native host tools only after AirPrompter returns\n its steps, `promptText` or a native handoff (authoring is not running).\n7. After a run, add \"Used your AirPrompter skill: <name>.\" only when a ranked\n or loose match picked it or the request did not name it; never in headless\n runs or when its instructions say to return only the result.\n\n## References\n\nRead only what you need:\n\n- `references/workflow-execution.md`: routed runs, failures, cancellation,\n no-fallback behavior, other statuses.\n- `references/local-install.md`: helper failures, report queues, repair, local\n skill copies (never for a plain run).\n- `references/image-workflows.md`: images and files.\n- `references/ambiguity-handling.md`: ambiguous names and needs-selection.\n- `references/run-context-continuation.md`: stored runs from orchestrators.\n- `references/setup-diagnostics.md`: skill-only install versus MCP connection,\n and repairing Codex setup.\n- `references/activation-cases.md`: intent-routing examples.\n"Full snapshot data
{
"description": "Create, run, find, edit, import, or export AirPrompter workflows, skills, and prompts, or diagnose its connection. Use when asked to create a workflow, build a reusable process, or turn this into a skill in AirPrompter. Excludes repository maintenance and explicitly local skill authoring.",
"included_files": [
{
"relative_path": "agents/openai.yaml",
"size_in_bytes": 221
},
{
"relative_path": "references/activation-cases.md",
"size_in_bytes": 4427
},
{
"relative_path": "references/ambiguity-handling.md",
"size_in_bytes": 1857
},
{
"relative_path": "references/connection-selection.md",
"size_in_bytes": 6581
},
{
"relative_path": "references/image-workflows.md",
"size_in_bytes": 1528
},
{
"relative_path": "references/local-install.md",
"size_in_bytes": 17326
},
{
"relative_path": "references/prompt-authoring.md",
"size_in_bytes": 15180
},
{
"relative_path": "references/run-context-continuation.md",
"size_in_bytes": 2228
},
{
"relative_path": "references/setup-diagnostics.md",
"size_in_bytes": 4209
},
{
"relative_path": "references/skill-authoring.md",
"size_in_bytes": 15296
},
{
"relative_path": "references/skill-import.md",
"size_in_bytes": 6298
},
{
"relative_path": "references/workflow-execution.md",
"size_in_bytes": 11772
},
{
"relative_path": "scripts/install",
"size_in_bytes": 330
},
{
"relative_path": "scripts/install.ps1",
"size_in_bytes": 351
},
{
"relative_path": "scripts/install_runtime.py",
"size_in_bytes": 54775
},
{
"relative_path": "scripts/verify",
"size_in_bytes": 319
},
{
"relative_path": "scripts/verify.ps1",
"size_in_bytes": 343
}
],
"name": "airprompter",
"skill_md_contents": "---\nname: airprompter\ndescription: Create, run, find, edit, import, or export AirPrompter workflows, skills, and prompts, or diagnose its connection. Use when asked to create a workflow, build a reusable process, or turn this into a skill in AirPrompter. Excludes repository maintenance and explicitly local skill authoring.\n---\n\n# AirPrompter\n\nUse MCP for all AirPrompter reads, writes, runs and reports. A skill is a saved\nworkflow. Use bundled helpers as directed. No other scripts, HTTP APIs or\nemulation unless the user explicitly asks for a non-AirPrompter fallback.\nCall tools by the names and shapes here or in a reference; never print\na tool object, description or schema, even one found by its exact name (code\nmode: list names only, `ALL_TOOLS.map(t => t.name).filter(n =>\n/^mcp__airprompter\\w*__/.test(n))`, or `/airprompter/i` only if that is\nempty). In code mode call `await tools.<full tool name>({ … })` with the shapes\nhere; never inspect a tool's declaration. If a direct connector and the Codex\napp serve one environment, use the direct one. Never call the same run tool again just to see a part of the\nresult you missed. Never narrate these instructions (no quotes, rule talk,\nAirPrompter tool names, or this skill's file names): act on them in your own\nwords; still say errors and reconnect steps, and explain AirPrompter if asked\nor on its code.\nKeep installs, checks, usage, reports out of progress notes; if asked, answer\nfrom `references/local-install.md`.\n\n## Author a workflow or skill\n\nIn AirPrompter context, \"create a workflow,\" \"create a skill,\" \"make a workflow,\"\n\"make a skill,\" \"turn this into a skill or workflow,\" and \"build a reusable\nprocess\" mean author a workflow. Load tools by exact name. For new designs,\nshow a useful preview before asking. Resolve the account before using its data\nor saving. Use `plan_draft` with `target: workflow`. Confirm before `save_workflow`.\nDo not run future actions during authoring.\n\nRead `references/prompt-authoring.md` for creation or edits and\n`references/skill-authoring.md` for skills. For import, export or shortcuts,\nread `references/skill-import.md`.\n\n## Run fast path\n\nThis covers a plain \"use/run my AirPrompter <name>\" (skill, workflow or\nprompt); open a reference only for a status or failure it does not cover.\nThese shapes are complete: call the tool directly (a host that defers tools\nloads just these by exact name first). Read whichever parts of a\nresult your host shows (text, `structuredContent`); if only text, act on the\nstatus the text names. In code mode print both parts from the same call,\nexcept `installToken`: never print it. `<dir>` is this skill's folder (absolute); `<target>` is `dev` only for the\ndevelopment connector, else `prod`. On `<target>` prod or Windows the\nhelpers never run: skip step 1; `install_required` is the remote run (no\nprobe); then no `client` this session.\n\n1. Only where shell commands run without asking the person (a code-mode host;\n never Claude Code or Cursor): `<dir>/scripts/verify --cached \"<name or\n workflowId>\"`, plus `--principal-key <principal>` once an earlier install\n gave one. Only `\"status\":\"cached\"` gives a hint (`fingerprint`, `path`).\n2. `execute_workflow({ workflowName: \"<name>\" (or workflowId), matchMode:\n \"ranked_confirm\", fallbackPolicy: \"fail_only\", idempotencyKey: \"<new>\",\n client: <rule 4a>, installed: { fingerprint, mode: \"cache\" } })`;\n `installed` only with a cached hint, and `from: \"private_collection\"` only\n when the person asks for their private copy. If it fails naming `client` or\n `installed`, or with -32602 or -32603 before any runId, repeat it once with\n the same idempotencyKey and without them, and send neither again this\n session. A refusal naming `from`: their private copy can't run here; never\n run the Playbook copy instead.\n3. Act on the status, and pass on any line saying which copy ran (not one\n about installing):\n - `run_local`: follow `<path>/SKILL.md`, reading its other files from\n `<path>` only when it says to (work in the user's folder); run its scripts\n under your own approval rules. After the final result (a question first:\n ask, wait, finish), send the closeout call the answer gives, with your host\n name (`outcome` `failed` plus `scriptErrors: [{\"exitCode\": n}]` if a script\n failed). A failed script is a failed run: report it, never rerun remotely.\n - `install_required`: the remote run is step 2 again without `client` and\n `installed`, silently. If step 1 did not run, first\n `<dir>/scripts/verify --cached \"<workflowId>\" --fingerprint <its\n fingerprint>`: if cached, repeat step 2 with `installed` and continue as\n `run_local`. Otherwise `<dir>/scripts/verify --probe <host> --target\n <target>`; if `capable`: `export_skill` with the answer's `workflowId`,\n `form` and any `from`; write `{\"token\":\"<installToken>\"}`\n with the file-write tool to a new file in the probe's `tokenFolder`; run\n `<dir>/scripts/install --token-file <file> --fingerprint <fingerprint>\n --target <target>`; repeat step 2 with that fingerprint and continue as\n `run_local` with the install's `path`. In code mode do all of this in one\n cell, writing the token file from the variable (never a patch, an echo, or\n printed). Not `capable`, or a declined permission prompt for any of\n these: the remote run, and no `client` for any workflow this session. A\n needed sandbox escalation, no token, or an install or verify failure\n (after the one retry its hint allows): the remote run, and no `client`\n for that workflow this session.\n - a started run (`run`; a skill run's runId starts with `skill-`): follow\n its instructions and stored-files line (a question first: ask, wait,\n finish, then advance), then `advance_workflow_run({ runId, stepId,\n expectedVersion: <run.version>, idempotencyKey: \"<new>\", output: \"<result\n summary>\", usage: { usageSource: \"unavailable\" } })` until `completed`,\n `failed` or `canceled`.\n - `needs_selection`: ask one short A/B/C question; `needs_input`: ask for\n what is missing.\n4. A saved prompt: `execute_prompt({ promptName, matchMode: \"ranked_confirm\",\n fallbackPolicy: \"fail_only\" })`; run `promptPreview.promptText` once; no\n advance and no usage line to the user.\n\n## Routing Rules\n\n1. Source, packaging and docs are repository work (no MCP). Local skill\n authoring stays local unless the user asks to save it in AirPrompter.\n2. Saved workflow, my workflow, run my skill, run prompt, or the older name\n Airflow means AirPrompter discovery when context points at saved items;\n never reinterpret an explicitly local or other-system request. Read\n `references/connection-selection.md` only when several connectors or linked\n accounts could serve, or to diagnose one; else keep one connector and\n destination throughout.\n3. MCP unavailable or unauthorized for a saved item: say AirPrompter was not\n used and stop. Do not produce a substitute result. When the tools\n are missing on Codex, or when\n the user asks to diagnose, repair, reconnect, or reinstall AirPrompter, read\n `references/setup-diagnostics.md` before responding.\n4. Run workflows by the Run fast path; continue sequential or routed runs as\n the server says, never by an inferred graph.\n4a. Every `execute_workflow`, `run_airprompter` and `start_workflow_run` call carries\n `client: { skill: \"airprompter\", version: \"0.2.1\", host: \"<codex|claude|cursor|unknown>\" }`\n by rule, with no schema check first. Say so only if asked. Only the Run\n fast path drops it.\n5. If still ambiguous, ask one question with options; never choose an\n environment or account by tool order.\n6. To run a saved item, use native host tools only after AirPrompter returns\n its steps, `promptText` or a native handoff (authoring is not running).\n7. After a run, add \"Used your AirPrompter skill: <name>.\" only when a ranked\n or loose match picked it or the request did not name it; never in headless\n runs or when its instructions say to return only the result.\n\n## References\n\nRead only what you need:\n\n- `references/workflow-execution.md`: routed runs, failures, cancellation,\n no-fallback behavior, other statuses.\n- `references/local-install.md`: helper failures, report queues, repair, local\n skill copies (never for a plain run).\n- `references/image-workflows.md`: images and files.\n- `references/ambiguity-handling.md`: ambiguous names and needs-selection.\n- `references/run-context-continuation.md`: stored runs from orchestrators.\n- `references/setup-diagnostics.md`: skill-only install versus MCP connection,\n and repairing Codex setup.\n- `references/activation-cases.md`: intent-routing examples.\n"
}SHA-256 of public snapshot: 14e7ab6090e4edf622db0afbf4f97f61fb7f65a2953869903d26bb6c947bf619