← FlowlinesCONTENT HISTORY

Update to Flowlines

Snapshot Oct 24, 2026 · 13:14 UTC · version 0.4.1

Collection source: downloaded plugin package. These snapshots do not have a confirmed matching collection source. Differences in file lists alone do not establish changes to the package.

WHAT CHANGED

Instructions updated for flowlines-weekly-review

Full description — Now says: only when the user explicitly asks for a weekly, monthly, or periodic review, a status report, or what changed over a period. Do not use for a single...

Observed in instructions or declared skills. Runtime behavior has not been tested.

Save Flowlines and check future changes in Following.

Up to 3 plugins free · no card required.

All fields and original values.

Full description

Before

Sep 30, 2026 · 23:10 UTC

when the user asks what changed, wants a weekly or monthly review, or asks for a status report on their agents.

After

Oct 24, 2026 · 13:14 UTC

only when the user explicitly asks for a weekly, monthly, or periodic review, a status report, or what changed over a period. Do not use for a single question about a metric, a session, a user, a release, or missing data; answer those di...

Skill instructions

Before

Sep 30, 2026 · 23:10 UTC

when the user asks what changed, wants a weekly or monthly review, or asks for a status report on their agents. 3. Focus agents, if the user names any. Otherwise cover every agent with activity. 5. `list_signals` for the window. For each...

After

Oct 24, 2026 · 13:14 UTC

only when the user explicitly asks for a weekly, monthly, or periodic review, a status report, or what changed over a period. Do not use for a single question about a metric, a session, a user, a release, or missing data; answer those di...

Compare saved observations

Download comparison JSON

SKILL.md text changes

− removed · + added

--- before
+++ after
@@ -1,12 +1,20 @@
 ---
 name: flowlines-weekly-review
-description: Produce a periodic review of a Flowlines namespace over the Flowlines MCP server - what changed since the last review, which signals fired, where outcomes moved, and what to pin for next time. Use when the user asks what changed, wants a weekly or monthly review, or asks for a status report on their agents.
+description: Produce a periodic review of a Flowlines namespace over the Flowlines MCP server - what changed since the last review, which signals fired, where outcomes moved, and what to pin for next time. Use only when the user explicitly asks for a weekly, monthly, or periodic review, a status report, or what changed over a period. Do not use for a single question about a metric, a session, a user, a release, or missing data; answer those directly with the MCP tools or use the matching Flowlines skill.
 ---
 
 # Flowlines weekly review
 
 Turn the Flowlines MCP tools into one repeatable review with a fixed shape, so successive reviews are comparable and nothing already known is re-derived.
 
+## Tool discipline
+
+- The Flowlines MCP server is the only data source for this skill. Do not open the Flowlines app, drive a browser, use computer-use, or search the web to find data, unless the user explicitly asks you to.
+- When the server cannot answer part of the question, say so in the answer and list it in `unmet_needs` of `report_outcome`. If the information lives in the Flowlines app, tell the user where to look; do not go there yourself.
+- Match the effort to the question. For a plain question, make the minimum calls (`get_workspace`, `get_context`, then the one or two tools that answer it), answer, and `report_outcome`. Run the full procedure below only when the user asks for a review or a status report, and skip `save_note` and evidence sessions unless the procedure runs in full or the user asks for them.
+- If authentication is required, or a generic MCP error (such as `-32603`) repeats on one read-only check, use `flowlines-doctor` when available for connection recovery before further data calls. Client diagnostics and native sign-in are allowed for this repair. After verification, resume the original request with the same namespace, range, and `user_intent`.
+- If `flowlines-doctor` is unavailable, use only the failing client's connection status and native reconnect action. Reconnect once only when that client reports an authentication problem; leave browser consent to the user, then verify with one read-only call. If the action is unavailable, give the user the reconnect step in their client and wait for completion. For an unexplained generic error or failed verification, stop data calls, give the user the error and a pointer to Flowlines support. Do not access Flowlines infrastructure or present unavailable data as empty.
+
 ## Conventions for every Flowlines tool call
 
 - Every tool takes `reason` (why this call) and `user_intent` (the user's goal, identical across the conversation). Set `user_intent` once, for example "Weekly review of the <namespace> namespace".
@@ -20,7 +28,7 @@
 
 1. The namespace. If `get_workspace` shows exactly one, use it silently.
 2. The review window. Default to the last 7 days. `get_changes_since` needs an ISO 8601 `since` timestamp and clamps to the trailing 90 days; take `since` from the previous review's pinned note when one exists, otherwise from the window.
-3. Focus agents, if the user names any. Otherwise cover every agent with activity.
+3. Focus MCP servers, if the user names any. Otherwise cover every MCP server with activity.
 
 ## Steps
 
@@ -28,8 +36,8 @@
 2. `list_notes`. Find the most recent note whose title starts with `Weekly review` and take `since` from its body. Titles carry the review date, so each period has its own note.
 3. `get_changes_since` with `since`. This returns session activity and outcomes in the window, signals that fired, and notes pinned. It replaces separate list calls.
 4. `aggregate_sessions` for the window with `group_by: ["agent"]` and metrics `session_count`, `user_count`, `success_count`, `failure_count`, `success_rate`, `total_cost_usd`, `unanalyzed_count`. Run it again with `group_by: ["day"]` for the trend. Never average rates across groups or days: a rate for any combination of groups is the summed `success_count` divided by the summed `success_count` plus `failure_count`. When a rate looks surprising, `get_metric_definition` for it before interpreting: denominators, missing-data rules, and sample floors differ by metric.
-5. `list_signals` for the window. For each signal that is new or grew, `get_signal` for its evidence. Group signals by agent and severity.
-6. For the largest movements, `aggregate_sessions` with `group_by: ["intent"]` on the affected agent to see which intents drive the change, then `list_sessions` filtered by `outcome: "unsuccessful"` or `user_feedback: "negative"` and open at most three sessions with `get_session` as evidence.
+5. `list_signals` for the window. For each signal that is new or grew, `get_signal` for its evidence. Group signals by MCP server and severity.
+6. For the largest movements, `aggregate_sessions` with `group_by: ["intent"]` on the affected MCP server to see which intents drive the change, then `list_sessions` filtered by `outcome: "unsuccessful"` or `user_feedback: "negative"` and open at most three sessions with `get_session` as evidence.
 7. Check identity coverage: `aggregate_sessions` with metric `session_count`, once with `include_unidentified: true` and once with `false`. The identified share is the second count divided by the first. `user_count` cannot show this gap because it never counts empty user ids. A low identified share means identity mapping is incomplete; report it rather than drawing user-level conclusions.
 8. Decide what to pin. `save_note` only for durable, verified findings that the next review must not rediscover: an evaluator artifact, an ingestion gap, a confirmed regression and its cause. Never put end-user personal data in a note.
 9. Save the continuation point. Pin one note titled `Weekly review <review date>`, for example `Weekly review 2026-09-02`, whose body records the review timestamp to use as the next `since`, the window covered, and the two or three headline numbers. The date in the title makes each period's note distinct; the server rejects a title it has seen recently regardless of the body, and notes cannot be edited. If the save is rejected as a duplicate, the period was already reviewed: `list_notes`, confirm the existing note covers this window, and reuse its timestamp. Pass `allow_duplicate: true` only when the existing note is for a different window and the title collided anyway. Confirm the save succeeded and the timestamp is recorded before reporting the review as complete.
@@ -42,8 +50,8 @@
 ```
 Namespace, window, review timestamp (use it as the next `since`)
 Headline: 3 bullets, each with the number and the change versus the previous window
-Outcomes by agent: table of sessions, users, success rate, cost, unanalysed
-Signals: new, grown, resolved - each with severity, agent, one-line evidence
+Outcomes by MCP server: table of sessions, users, success rate, cost, unanalysed
+Signals: new, grown, resolved - each with severity, MCP server, one-line evidence
 Notable intents: the intents behind the largest movements
 Data quality: unanalysed share, unidentified users, context status, anything pinned as a caveat
 Pinned this review: titles of notes saved, including the dated review note and its timestamp
Full technical diff · 2 changed fields

changed /description

BEFORE
"Produce a periodic review of a Flowlines namespace over the Flowlines MCP server - what changed since the last review, which signals fired, where outcomes moved, and what to pin for next time. Use when the user asks what changed, wants a weekly or monthly review, or asks for a status report on their agents."
AFTER
"Produce a periodic review of a Flowlines namespace over the Flowlines MCP server - what changed since the last review, which signals fired, where outcomes moved, and what to pin for next time. Use only when the user explicitly asks for a weekly, monthly, or periodic review, a status report, or what changed over a period. Do not use for a single question about a metric, a session, a user, a release, or missing data; answer those directly with the MCP tools or use the matching Flowlines skill."

changed /skill_md_contents

BEFORE
"---\nname: flowlines-weekly-review\ndescription: Produce a periodic review of a Flowlines namespace over the Flowlines MCP server - what changed since the last review, which signals fired, where outcomes moved, and what to pin for next time. Use when the user asks what changed, wants a weekly or monthly review, or asks for a status report on their agents.\n---\n\n# Flowlines weekly review\n\nTurn the Flowlines MCP tools into one repeatable review with a fixed shape, so successive reviews are comparable and nothing already known is re-derived.\n\n## Conventions for every Flowlines tool call\n\n- Every tool takes `reason` (why this call) and `user_intent` (the user's goal, identical across the conversation). Set `user_intent` once, for example \"Weekly review of the <namespace> namespace\".\n- Start with `get_workspace` once, pick the namespace, then `get_context` for that namespace. Read its pinned notes and glossary before analysing: notes record known caveats and confirmed findings.\n- Prefer aggregates over transcripts. Open a session only as evidence for a claim, quote the minimum, and mask names, emails, and other identifiers.\n- End with `report_outcome` as the last tool call, before the final answer, with `unmet_needs` for anything the server could not provide.\n\n## Inputs\n\nAsk for, or infer, three things:\n\n1. The namespace. If `get_workspace` shows exactly one, use it silently.\n2. The review window. Default to the last 7 days. `get_changes_since` needs an ISO 8601 `since` timestamp and clamps to the trailing 90 days; take `since` from the previous review's pinned note when one exists, otherwise from the window.\n3. Focus agents, if the user names any. Otherwise cover every agent with activity.\n\n## Steps\n\n1. `get_context` with `overview_range` matching the window (`7d` for a weekly review, `30d` for a monthly one). Record the context status; if it is `not_configured`, `gathering`, or `failed`, say so and continue without it.\n2. `list_notes`. Find the most recent note whose title starts with `Weekly review` and take `since` from its body. Titles carry the review date, so each period has its own note.\n3. `get_changes_since` with `since`. This returns session activity and outcomes in the window, signals that fired, and notes pinned. It replaces separate list calls.\n4. `aggregate_sessions` for the window with `group_by: [\"agent\"]` and metrics `session_count`, `user_count`, `success_count`, `failure_count`, `success_rate`, `total_cost_usd`, `unanalyzed_count`. Run it again with `group_by: [\"day\"]` for the trend. Never average rates across groups or days: a rate for any combination of groups is the summed `success_count` divided by the summed `success_count` plus `failure_count`. When a rate looks surprising, `get_metric_definition` for it before interpreting: denominators, missing-data rules, and sample floors differ by metric.\n5. `list_signals` for the window. For each signal that is new or grew, `get_signal` for its evidence. Group signals by agent and severity.\n6. For the largest movements, `aggregate_sessions` with `group_by: [\"intent\"]` on the affected agent to see which intents drive the change, then `list_sessions` filtered by `outcome: \"unsuccessful\"` or `user_feedback: \"negative\"` and open at most three sessions with `get_session` as evidence.\n7. Check identity coverage: `aggregate_sessions` with metric `session_count`, once with `include_unidentified: true` and once with `false`. The identified share is the second count divided by the first. `user_count` cannot show this gap because it never counts empty user ids. A low identified share means identity mapping is incomplete; report it rather than drawing user-level conclusions.\n8. Decide what to pin. `save_note` only for durable, verified findings that the next review must not rediscover: an evaluator artifact, an ingestion gap, a confirmed regression and its cause. Never put end-user personal data in a note.\n9. Save the continuation point. Pin one note titled `Weekly review <review date>`, for example `Weekly review 2026-09-02`, whose body records the review timestamp to use as the next `since`, the window covered, and the two or three headline numbers. The date in the title makes each period's note distinct; the server rejects a title it has seen recently regardless of the body, and notes cannot be edited. If the save is rejected as a duplicate, the period was already reviewed: `list_notes`, confirm the existing note covers this window, and reuse its timestamp. Pass `allow_duplicate: true` only when the existing note is for a different window and the title collided anyway. Confirm the save succeeded and the timestamp is recorded before reporting the review as complete.\n10. `report_outcome`.\n\n## Output shape\n\nUse this structure every time so reviews line up week over week:\n\n```\nNamespace, window, review timestamp (use it as the next `since`)\nHeadline: 3 bullets, each with the number and the change versus the previous window\nOutcomes by agent: table of sessions, users, success rate, cost, unanalysed\nSignals: new, grown, resolved - each with severity, agent, one-line evidence\nNotable intents: the intents behind the largest movements\nData quality: unanalysed share, unidentified users, context status, anything pinned as a caveat\nPinned this review: titles of notes saved, including the dated review note and its timestamp\nOpen questions: what the server could not answer (also sent as unmet_needs)\n```\n\nReport numbers with their metric definitions in mind: an `unanalyzed_count` above a few percent makes `success_rate` provisional, and a rate below the metric's sample floor is not a trend. Say when a comparison is against a partial previous window.\n\n## Do not\n\n- Do not page through `list_sessions` to build totals; `aggregate_sessions` exists for that.\n- Do not average rates across groups; recompute from summed counts.\n- Do not quote prompts or responses beyond the minimum needed to support a finding.\n- Do not pin conversation scratch, hunches, or per-user observations.\n- Do not skip `report_outcome`, including when the review is empty or blocked.\n"
AFTER
"---\nname: flowlines-weekly-review\ndescription: Produce a periodic review of a Flowlines namespace over the Flowlines MCP server - what changed since the last review, which signals fired, where outcomes moved, and what to pin for next time. Use only when the user explicitly asks for a weekly, monthly, or periodic review, a status report, or what changed over a period. Do not use for a single question about a metric, a session, a user, a release, or missing data; answer those directly with the MCP tools or use the matching Flowlines skill.\n---\n\n# Flowlines weekly review\n\nTurn the Flowlines MCP tools into one repeatable review with a fixed shape, so successive reviews are comparable and nothing already known is re-derived.\n\n## Tool discipline\n\n- The Flowlines MCP server is the only data source for this skill. Do not open the Flowlines app, drive a browser, use computer-use, or search the web to find data, unless the user explicitly asks you to.\n- When the server cannot answer part of the question, say so in the answer and list it in `unmet_needs` of `report_outcome`. If the information lives in the Flowlines app, tell the user where to look; do not go there yourself.\n- Match the effort to the question. For a plain question, make the minimum calls (`get_workspace`, `get_context`, then the one or two tools that answer it), answer, and `report_outcome`. Run the full procedure below only when the user asks for a review or a status report, and skip `save_note` and evidence sessions unless the procedure runs in full or the user asks for them.\n- If authentication is required, or a generic MCP error (such as `-32603`) repeats on one read-only check, use `flowlines-doctor` when available for connection recovery before further data calls. Client diagnostics and native sign-in are allowed for this repair. After verification, resume the original request with the same namespace, range, and `user_intent`.\n- If `flowlines-doctor` is unavailable, use only the failing client's connection status and native reconnect action. Reconnect once only when that client reports an authentication problem; leave browser consent to the user, then verify with one read-only call. If the action is unavailable, give the user the reconnect step in their client and wait for completion. For an unexplained generic error or failed verification, stop data calls, give the user the error and a pointer to Flowlines support. Do not access Flowlines infrastructure or present unavailable data as empty.\n\n## Conventions for every Flowlines tool call\n\n- Every tool takes `reason` (why this call) and `user_intent` (the user's goal, identical across the conversation). Set `user_intent` once, for example \"Weekly review of the <namespace> namespace\".\n- Start with `get_workspace` once, pick the namespace, then `get_context` for that namespace. Read its pinned notes and glossary before analysing: notes record known caveats and confirmed findings.\n- Prefer aggregates over transcripts. Open a session only as evidence for a claim, quote the minimum, and mask names, emails, and other identifiers.\n- End with `report_outcome` as the last tool call, before the final answer, with `unmet_needs` for anything the server could not provide.\n\n## Inputs\n\nAsk for, or infer, three things:\n\n1. The namespace. If `get_workspace` shows exactly one, use it silently.\n2. The review window. Default to the last 7 days. `get_changes_since` needs an ISO 8601 `since` timestamp and clamps to the trailing 90 days; take `since` from the previous review's pinned note when one exists, otherwise from the window.\n3. Focus MCP servers, if the user names any. Otherwise cover every MCP server with activity.\n\n## Steps\n\n1. `get_context` with `overview_range` matching the window (`7d` for a weekly review, `30d` for a monthly one). Record the context status; if it is `not_configured`, `gathering`, or `failed`, say so and continue without it.\n2. `list_notes`. Find the most recent note whose title starts with `Weekly review` and take `since` from its body. Titles carry the review date, so each period has its own note.\n3. `get_changes_since` with `since`. This returns session activity and outcomes in the window, signals that fired, and notes pinned. It replaces separate list calls.\n4. `aggregate_sessions` for the window with `group_by: [\"agent\"]` and metrics `session_count`, `user_count`, `success_count`, `failure_count`, `success_rate`, `total_cost_usd`, `unanalyzed_count`. Run it again with `group_by: [\"day\"]` for the trend. Never average rates across groups or days: a rate for any combination of groups is the summed `success_count` divided by the summed `success_count` plus `failure_count`. When a rate looks surprising, `get_metric_definition` for it before interpreting: denominators, missing-data rules, and sample floors differ by metric.\n5. `list_signals` for the window. For each signal that is new or grew, `get_signal` for its evidence. Group signals by MCP server and severity.\n6. For the largest movements, `aggregate_sessions` with `group_by: [\"intent\"]` on the affected MCP server to see which intents drive the change, then `list_sessions` filtered by `outcome: \"unsuccessful\"` or `user_feedback: \"negative\"` and open at most three sessions with `get_session` as evidence.\n7. Check identity coverage: `aggregate_sessions` with metric `session_count`, once with `include_unidentified: true` and once with `false`. The identified share is the second count divided by the first. `user_count` cannot show this gap because it never counts empty user ids. A low identified share means identity mapping is incomplete; report it rather than drawing user-level conclusions.\n8. Decide what to pin. `save_note` only for durable, verified findings that the next review must not rediscover: an evaluator artifact, an ingestion gap, a confirmed regression and its cause. Never put end-user personal data in a note.\n9. Save the continuation point. Pin one note titled `Weekly review <review date>`, for example `Weekly review 2026-09-02`, whose body records the review timestamp to use as the next `since`, the window covered, and the two or three headline numbers. The date in the title makes each period's note distinct; the server rejects a title it has seen recently regardless of the body, and notes cannot be edited. If the save is rejected as a duplicate, the period was already reviewed: `list_notes`, confirm the existing note covers this window, and reuse its timestamp. Pass `allow_duplicate: true` only when the existing note is for a different window and the title collided anyway. Confirm the save succeeded and the timestamp is recorded before reporting the review as complete.\n10. `report_outcome`.\n\n## Output shape\n\nUse this structure every time so reviews line up week over week:\n\n```\nNamespace, window, review timestamp (use it as the next `since`)\nHeadline: 3 bullets, each with the number and the change versus the previous window\nOutcomes by MCP server: table of sessions, users, success rate, cost, unanalysed\nSignals: new, grown, resolved - each with severity, MCP server, one-line evidence\nNotable intents: the intents behind the largest movements\nData quality: unanalysed share, unidentified users, context status, anything pinned as a caveat\nPinned this review: titles of notes saved, including the dated review note and its timestamp\nOpen questions: what the server could not answer (also sent as unmet_needs)\n```\n\nReport numbers with their metric definitions in mind: an `unanalyzed_count` above a few percent makes `success_rate` provisional, and a rate below the metric's sample floor is not a trend. Say when a comparison is against a partial previous window.\n\n## Do not\n\n- Do not page through `list_sessions` to build totals; `aggregate_sessions` exists for that.\n- Do not average rates across groups; recompute from summed counts.\n- Do not quote prompts or responses beyond the minimum needed to support a finding.\n- Do not pin conversation scratch, hunches, or per-user observations.\n- Do not skip `report_outcome`, including when the review is empty or blocked.\n"
Full snapshot data
{
  "description": "Produce a periodic review of a Flowlines namespace over the Flowlines MCP server - what changed since the last review, which signals fired, where outcomes moved, and what to pin for next time. Use only when the user explicitly asks for a weekly, monthly, or periodic review, a status report, or what changed over a period. Do not use for a single question about a metric, a session, a user, a release, or missing data; answer those directly with the MCP tools or use the matching Flowlines skill.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 229
    }
  ],
  "name": "flowlines-weekly-review",
  "skill_md_contents": "---\nname: flowlines-weekly-review\ndescription: Produce a periodic review of a Flowlines namespace over the Flowlines MCP server - what changed since the last review, which signals fired, where outcomes moved, and what to pin for next time. Use only when the user explicitly asks for a weekly, monthly, or periodic review, a status report, or what changed over a period. Do not use for a single question about a metric, a session, a user, a release, or missing data; answer those directly with the MCP tools or use the matching Flowlines skill.\n---\n\n# Flowlines weekly review\n\nTurn the Flowlines MCP tools into one repeatable review with a fixed shape, so successive reviews are comparable and nothing already known is re-derived.\n\n## Tool discipline\n\n- The Flowlines MCP server is the only data source for this skill. Do not open the Flowlines app, drive a browser, use computer-use, or search the web to find data, unless the user explicitly asks you to.\n- When the server cannot answer part of the question, say so in the answer and list it in `unmet_needs` of `report_outcome`. If the information lives in the Flowlines app, tell the user where to look; do not go there yourself.\n- Match the effort to the question. For a plain question, make the minimum calls (`get_workspace`, `get_context`, then the one or two tools that answer it), answer, and `report_outcome`. Run the full procedure below only when the user asks for a review or a status report, and skip `save_note` and evidence sessions unless the procedure runs in full or the user asks for them.\n- If authentication is required, or a generic MCP error (such as `-32603`) repeats on one read-only check, use `flowlines-doctor` when available for connection recovery before further data calls. Client diagnostics and native sign-in are allowed for this repair. After verification, resume the original request with the same namespace, range, and `user_intent`.\n- If `flowlines-doctor` is unavailable, use only the failing client's connection status and native reconnect action. Reconnect once only when that client reports an authentication problem; leave browser consent to the user, then verify with one read-only call. If the action is unavailable, give the user the reconnect step in their client and wait for completion. For an unexplained generic error or failed verification, stop data calls, give the user the error and a pointer to Flowlines support. Do not access Flowlines infrastructure or present unavailable data as empty.\n\n## Conventions for every Flowlines tool call\n\n- Every tool takes `reason` (why this call) and `user_intent` (the user's goal, identical across the conversation). Set `user_intent` once, for example \"Weekly review of the <namespace> namespace\".\n- Start with `get_workspace` once, pick the namespace, then `get_context` for that namespace. Read its pinned notes and glossary before analysing: notes record known caveats and confirmed findings.\n- Prefer aggregates over transcripts. Open a session only as evidence for a claim, quote the minimum, and mask names, emails, and other identifiers.\n- End with `report_outcome` as the last tool call, before the final answer, with `unmet_needs` for anything the server could not provide.\n\n## Inputs\n\nAsk for, or infer, three things:\n\n1. The namespace. If `get_workspace` shows exactly one, use it silently.\n2. The review window. Default to the last 7 days. `get_changes_since` needs an ISO 8601 `since` timestamp and clamps to the trailing 90 days; take `since` from the previous review's pinned note when one exists, otherwise from the window.\n3. Focus MCP servers, if the user names any. Otherwise cover every MCP server with activity.\n\n## Steps\n\n1. `get_context` with `overview_range` matching the window (`7d` for a weekly review, `30d` for a monthly one). Record the context status; if it is `not_configured`, `gathering`, or `failed`, say so and continue without it.\n2. `list_notes`. Find the most recent note whose title starts with `Weekly review` and take `since` from its body. Titles carry the review date, so each period has its own note.\n3. `get_changes_since` with `since`. This returns session activity and outcomes in the window, signals that fired, and notes pinned. It replaces separate list calls.\n4. `aggregate_sessions` for the window with `group_by: [\"agent\"]` and metrics `session_count`, `user_count`, `success_count`, `failure_count`, `success_rate`, `total_cost_usd`, `unanalyzed_count`. Run it again with `group_by: [\"day\"]` for the trend. Never average rates across groups or days: a rate for any combination of groups is the summed `success_count` divided by the summed `success_count` plus `failure_count`. When a rate looks surprising, `get_metric_definition` for it before interpreting: denominators, missing-data rules, and sample floors differ by metric.\n5. `list_signals` for the window. For each signal that is new or grew, `get_signal` for its evidence. Group signals by MCP server and severity.\n6. For the largest movements, `aggregate_sessions` with `group_by: [\"intent\"]` on the affected MCP server to see which intents drive the change, then `list_sessions` filtered by `outcome: \"unsuccessful\"` or `user_feedback: \"negative\"` and open at most three sessions with `get_session` as evidence.\n7. Check identity coverage: `aggregate_sessions` with metric `session_count`, once with `include_unidentified: true` and once with `false`. The identified share is the second count divided by the first. `user_count` cannot show this gap because it never counts empty user ids. A low identified share means identity mapping is incomplete; report it rather than drawing user-level conclusions.\n8. Decide what to pin. `save_note` only for durable, verified findings that the next review must not rediscover: an evaluator artifact, an ingestion gap, a confirmed regression and its cause. Never put end-user personal data in a note.\n9. Save the continuation point. Pin one note titled `Weekly review <review date>`, for example `Weekly review 2026-09-02`, whose body records the review timestamp to use as the next `since`, the window covered, and the two or three headline numbers. The date in the title makes each period's note distinct; the server rejects a title it has seen recently regardless of the body, and notes cannot be edited. If the save is rejected as a duplicate, the period was already reviewed: `list_notes`, confirm the existing note covers this window, and reuse its timestamp. Pass `allow_duplicate: true` only when the existing note is for a different window and the title collided anyway. Confirm the save succeeded and the timestamp is recorded before reporting the review as complete.\n10. `report_outcome`.\n\n## Output shape\n\nUse this structure every time so reviews line up week over week:\n\n```\nNamespace, window, review timestamp (use it as the next `since`)\nHeadline: 3 bullets, each with the number and the change versus the previous window\nOutcomes by MCP server: table of sessions, users, success rate, cost, unanalysed\nSignals: new, grown, resolved - each with severity, MCP server, one-line evidence\nNotable intents: the intents behind the largest movements\nData quality: unanalysed share, unidentified users, context status, anything pinned as a caveat\nPinned this review: titles of notes saved, including the dated review note and its timestamp\nOpen questions: what the server could not answer (also sent as unmet_needs)\n```\n\nReport numbers with their metric definitions in mind: an `unanalyzed_count` above a few percent makes `success_rate` provisional, and a rate below the metric's sample floor is not a trend. Say when a comparison is against a partial previous window.\n\n## Do not\n\n- Do not page through `list_sessions` to build totals; `aggregate_sessions` exists for that.\n- Do not average rates across groups; recompute from summed counts.\n- Do not quote prompts or responses beyond the minimum needed to support a finding.\n- Do not pin conversation scratch, hunches, or per-user observations.\n- Do not skip `report_outcome`, including when the review is empty or blocked.\n"
}

SHA-256 of public snapshot: a242ca6d0fac0825a84995bc0b9493d2c006329482f9c2467e2132709dfabe68