{"id":16694,"plugin_id":"plugins_6a630817fcdc8191a6e0114bbd61bc33","kind":"skill","collection_source":null,"comparison_source":null,"observed_at":"2026-09-30T23:13:39.813Z","digest":"697f5e839f51517f313cd2201897db58373f40e476e6230230763e61d31e1ffd","against":null,"payload":{"description":"Learn reusable working methods from selected experience and propose handbook practices, checks, Playbooks, templates, or coordinated improvements. Use the current conversation, exact selected sessions, or explicitly supplied work evidence. Return editable Shape Handbook requests without writing files or applying proposals automatically. Do not use for research or analysis unrelated to working methods.","included_files":[{"relative_path":"agents/openai.yaml","size_in_bytes":251},{"relative_path":"references/CONTRACT.md","size_in_bytes":15762},{"relative_path":"references/methodology.md","size_in_bytes":10863}],"name":"learn-from-work","skill_md_contents":"---\nname: learn-from-work\ndescription: Learn reusable working methods from selected experience and propose handbook practices, checks, Playbooks, templates, or coordinated improvements. Use the current conversation, exact selected sessions, or explicitly supplied work evidence. Return editable Shape Handbook requests without writing files or applying proposals automatically. Do not use for research or analysis unrelated to working methods.\n---\n\n# Learn From Work\n\nBefore applying this skill, read the full [Trellis methodology](references/methodology.md).\n\nFind what selected experience suggests preserving or improving about how an area\nworks. Recommend the form that serves the lesson; do not force every useful\ndistinction, check, or template into a Playbook. Learning produces proposals,\nnot handbook changes or automatic calls to Shape Handbook.\n\n## Select the evidence\n\nUse the current conversation by default, exact sessions selected by the user and\nretrievable through the host, or explicitly selected work evidence such as artifacts\nor an account of an episode. Do not enumerate unselected history, guess identifiers,\ninspect adjacent conversations, or search unrelated folders to fill gaps. A selected\nsession does not authorize reading every file or service mentioned in it.\n\nReport unavailable, ambiguous, or partial sources individually and continue with\nusable evidence. State which sources you examined. If host limits require batches,\nidentify them rather than silently sampling or omitting sources. Reconstruct each\nepisode independently before comparing patterns. One source cannot establish what\nan unavailable source contained.\n\nWhen a handbook or surrounding area is selected, inspect relevant existing method\nand shared guidance to judge fit and avoid duplication. Resolve workspace and area\nfrom the request, project instructions, and README. A handbook is `<area>/handbook/`;\nits parent owns the area. Its README may declare the workspace through:\n\n```html\n<!-- trellis:handbook {\"format\":1,\"workspace\":\"../..\"} -->\n```\n\nThe path is relative to the handbook; a workspace-root handbook uses `..`. Verify\nphysical scope against the selected workspace, which need not be a Git repository.\nDo not use the current directory, a marker, or symlinks to expand ambiguous scope.\nResolve multiple selected handbooks separately. Root and nested guidance can both\napply; ancestry does not settle conflicting rules. Avoid unrelated areas.\n\nA destination need not be chosen before a useful proposal can be made. State any\nunresolved handbook choice rather than inventing it. Ask only when the answer\nmaterially changes the proposed method or its fit.\n\n## Reconstruct before drawing lessons\n\nIdentify the starting situation, intended outcome, context, constraints, decisions\nand cues, actions, results, verification, failures, and recovery relevant to the\nproposed lesson. Distinguish direct observations, the user's account, stated intent,\nand inference. An artifact records its contents; it does not establish an unobserved\nprocess or prove that its author followed a particular method.\n\nLook for useful turning points: a hidden constraint, correction, meaningful\ndistinction, failed approach, check that exposes an important distinction,\nrecurring difficulty, or recovery.\nPrefer reasoning that would change future work over a chronology of tool calls.\n\nAn unfinished attempt can suggest a method without demonstrating its outcome.\nSuccess does not establish which actions caused it or whether the method generalizes.\nConsider another explanation, such as a reviewer's correction or a favorable input,\nand a situation where the lesson would not help. Keep these interpretations distinct\nfrom observed evidence; do not invent a counterexample or claim an unperformed\ncomparison. Current clarification can shape a proposal without rewriting the selected\nevidence. Repetition alone does not turn a pattern into adopted policy.\n\nWhen a missing fact or interpretation would change the lesson, ask a focused\nquestion before finalizing that recommendation. Ground it in the selected episode:\nwhat prompted a correction, which outcome mattered, what happened afterward, or\nwhere the proposed method would not fit. Ask only questions whose plausible answers\nwould change the advice, and use answers already present in the sources. Do not\nhand the analysis back as a generic reflection questionnaire.\n\nWait for the answer on that dependency while continuing supported independent\nanalysis. Use the reply to revise, narrow, or discard the candidate lesson, keeping\nthe user's present account distinct from inspected historical evidence. If the\nanswer is unknown or declined, state what remains tentative; do not keep asking\nor convert uncertainty into a causal claim. Enough evidence needs no extra interview.\n\n## Choose useful handbook improvements\n\nConsider what would help a future contributor:\n\n- **Practice:** a working distinction, convention, approach, or interpretation that\n  changes recurring decisions.\n- **Check:** an independently applicable expectation with concrete verification\n  and meaningful outcomes.\n- **Playbook:** a coherent repeatable task with a clear trigger, judgment, method,\n  outcome, completion evidence, and meaningful recovery or stop conditions.\n- **Template:** an adaptable structure or substantive prompts that improve\n  repeated authoring.\n- **Coordinated refinement:** an existing handbook concern that needs clearer\n  ownership, links, scope, organization, or changes across several artifact types.\n\nFor a candidate lesson, identify a recognizable cue, the decision it would change,\nwhy that change might help, its applicability and exceptions, and evidence that\ncould test it. Use these to assess usefulness, confidence, and maintenance cost;\nthey do not require separate headings. Prefer refining an existing owner when it\nalready covers the concern. Keep connected changes together and independently\nuseful proposals separate. Finding nothing worth preserving is a valid result.\n\nPropose how to improve the working method while leaving the actual work in its\nexisting files. Judge mixed material statement by statement: a current task list,\nresult, or frequently consulted fact is not itself a reusable method. Link to the\nauthoritative work files when needed rather than copying their content. Replace\nincidental names, dates, values, and paths with parameters. Never retain secrets,\nunnecessary personal data, transcripts, or session identifiers in proposed handbook\nartifacts.\n\n## Return self-contained, editable requests\n\nLead with the useful conclusion: what could change next time, when it would help,\nwhat supports it, and what remains uncertain. A question alone does not complete\nthe reflection after an answer is available. Explain a supported no-change result;\ndo not invent a lesson to fill a quota. For each worthwhile proposal, then return\na request the user can edit, combine, ignore, or choose, for example:\n\n```text\nUse $shape-handbook to refine [selected handbook, if known] so future [work] can\n[working improvement]. Preserve [practice or distinction], define [check and\nverification] where useful, and update [relevant template or existing owner].\nThe selected evidence shows ... It leaves ... unresolved. Keep ... visibly proposed.\n```\n\nInclude enough context to shape the method without reopening the source conversation:\napplicability, proposed change, useful decisions and exceptions, verification,\nauthoritative sources or existing material to revise, observed evidence and its limits,\nand remaining dependencies. Distinguish inspected results from interpretations and\nproposed work. Keep consequential inferred rules proposed until adopted. Do not invent\npolicy, credentials, endpoints, authority, or proof of effectiveness.\n\nWhen proposing a Playbook, explicitly describe the procedure to preserve and its\noutcome in the request. The user must choose that proposal or independently request\nthe procedure before it is created. A learning recommendation is not itself that\nauthorization. Suggest supporting scripts only when materially useful and explain\ntheir proposed role within the workspace; do not generate them.\n\n## Read-only boundary\n\nPrefer direct file reads; when shell reads are needed, use a non-login shell to\navoid cache-writing login hooks.\n\nPerform no filesystem writes, including reports, transcripts, temporary files,\ncaches, or generated artifacts. Return proposals in the response and never invoke\nShape Handbook automatically. Do not execute described work, Playbooks, scripts,\nor tests; install dependencies; edit `AGENTS.md`; initialize Git; stage, commit,\npush, publish; or communicate with others. Learning a method neither performs it\nnor grants authority to perform it later. Another agent using the proposal retains\nits separately authorized task permissions; neither operation inherits the other's\nauthority, and delegation cannot bypass this skill's boundaries.\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}