AGENT SKILL

tk-studio-research

Run a chartered research pass — survey the topics a charter scopes, grade every finding by evidence, land the findings in the run workspace, and emit observation events for recommendation-worthy ones. Use when the user says "run the ecosystem watch", "tk research", or a research job's invoke-skill directive names this skill.

License 许可
MIT
最近更新
2026-11-10
Stars
★ 0
Forks
0

01 前言 Front Matter

name
tk-studio-research
description
Run a chartered research pass — survey the topics a charter scopes, grade every finding by evidence, land the findings in the run workspace, and emit observation events for recommendation-worthy ones. Use when the user says "run the ecosystem watch", "tk research", or a research job's invoke-skill directive names this skill.

02 资源文件 1

03 SKILL.md 正文

tk-studio-research

The research half of the evolve loop's input engine (ST-6.4, AD-8: observe and log only — findings feed recommendations, they never trigger changes). The deterministic core keeps a run honest at both ends: no charter, no run; no evidence grade, no finding.

Behavior (both modes)

  1. Charter first. The scope arrives in the payload (a job run carries it at target.payload.charter); validate it before any research:

    uv run "${CLAUDE_PLUGIN_ROOT}/lib/research.py" charter --charter '{"topics": ["..."], "sources": ["..."]}'
    

    Topics and sources are required and non-empty — an unscoped "go research" ends blocked (headless) or asks once for scope (attended).

  2. Seed the run (when anchors matter). If the run's findings will cite or correct standing assumptions, author a per-run seed and land it through the core before researching (knowledge.schema.json shape — tier: seed, the verbatim supersede header, [SEED-<run-id>-A<n>] anchors, inherits: listing any project-spine anchors reused):

    uv run "${CLAUDE_PLUGIN_ROOT}/lib/research.py" seed --directory <root> --run-id <rid> --file <seed.md>
    

    The core validates via lib/knowledge.py, verifies inherits against the project spine when one exists, and reports a missing spine without blocking — aid, not gate. A run that cites no anchors needs no seed.

  3. Research within budget. Survey the chartered topics through the chartered sources only. When running as a job (an invoke-skill directive from tk-studio-job), call jobrun.py account --turns 1 every iteration and stop the moment it answers within_budget: false. Cap output at the charter's max_findings — depth beats sprawl.

  4. Grade everything. Every finding carries an evidence grade and its trail: A primary source, verified directly; B multiple independent sources agree; C single credible source; D inference or speculation. A finding worth acting on carries a recommendation with a concrete action. A finding about a seeded assumption cites it: optional anchor = a bare anchor id from the seed's available set (defined + inherited) — dangling citations are named rejections.

  5. Land and emit through the core (never hand-write the artifacts):

    uv run "${CLAUDE_PLUGIN_ROOT}/lib/research.py" record --directory <root> --run-id <rid> --findings-file <findings.json>
    

    The core validates the grades (and any anchor citations against the seed), lands findings.json + findings.md in the run workspace, reports project-spine presence, and emits one observation event per recommendation-carrying finding (source research-job) — this surface's own events; the job wrapper's job-run event stays the wrapper's (AD-12). Then end the run via jobrun.py finish.

  6. Attended: summarize the graded findings and where they landed. Headless: a missing/invalid charter ends blocked, a budget hit ends partial with the guard named — never a prompt (AD-11). End headless runs with the status block:

    {"status": "complete", "intent": "tk-studio-research", "artifacts": ["findings.json", "findings.md"], "reason": null}
    

If the deterministic core is unrunnable — a tool call denied by permissions, uv/python unavailable — end blocked with the status block naming the unrunnable core as reason: never a question, never a headless run that ends without the block (AD-11).

The spine flavor (project scope)

A chartered research run may be charged with authoring or growing the project research spine — the broad once-per-project assumption map that seeds inherit. Same discipline, different scope: research the project's actual systems and invariants within the charter, then land through the core (never hand-write into the store):

uv run "${CLAUDE_PLUGIN_ROOT}/lib/research.py" spine --directory <root> --run-id <rid> --file <spine.md>

The artifact carries tier: spine, the verbatim supersede header, and [SPINE-A<n>] anchors; it lands at ~/.tk-studio/projects/<key>/knowledge/spine.md (per-user store, D2 — provisional knowledge never lands on the project VCS). Anchor ids are append-only: re-authoring must keep every existing id, or the core refuses — queued deltas cite identity.

Rules

  • Observe and log is the boundary (AD-8): findings and observations only — never draft skills, edit configs, or open PRs from a research run.
  • Ungraded or evidence-free claims don't land: grades A–C require a non-empty evidence trail; only D may stand bare, marked speculative.
  • Recommendation-worthy = carries a concrete recommendation.action; that is what emits an observation, nothing else.
  • All paths resolve through ${CLAUDE_PLUGIN_ROOT}.

04 同仓库更多 Skills

thkennedy
thkennedy/tk-studio

tk-studio-track

Tracking verbs on the bound Jira project — `block` moves a story the loop cannot progress, and each dependent story, to "Blocked - Awaiting User" with a comment that @mentions the owner (or, with `--engagement`, one engagement's questions issue, which `close` later moves to "Done"); `reply` posts the owner's answer (or "Done.") with the owner's host-only token and moves the issue to "Block Resolved"; `validate` checks that resolution and passes the issue back to work or re-blocks it with a follow-up. Use when the user says "block this story", "tk track block", "put it in the owner's inbox", "reply to the blocked issue", "validate the resolution", or a supervisor or engagement payload names this skill.

★ 0
thkennedy
thkennedy/tk-studio

tk-studio-session

Session discipline as a surface — land a boundary handoff (optionally carrying knowledge deltas) and end the session, or resume a fresh session from a run workspace alone. Use when the user says "hand off this run", "tk session", "resume the run", or a boundary/budget trigger says the session must split.

★ 0
thkennedy
thkennedy/tk-studio

tk-studio-report

One-verb human defect report — "the skill did the wrong thing" — recorded as a report event in the local measurement ledger. Use when the user says "tk report", "report a defect", "that skill misbehaved", or "log a studio issue".

★ 0
thkennedy
thkennedy/tk-studio

tk-studio-playable-judge

The playable-verify leg's screenshot judge (Story 34.3) — reads only a playable run's report.json and its frames, answers the fixed rubric (scene loaded, player visible and framed, no magenta or error overlay, HUD reads, the named moments shown) with pass or fail and findings each naming a frame and a rubric line. Driven headless by the supervisor's host leg on the project's sandbox tier; use when the payload names a playable `media` folder, or the user says "judge the playable frames".

★ 0
thkennedy
thkennedy/tk-studio

tk-studio-plan-sync

Planning adapter — canonicalize BMad-native planning artifacts (normalize pass, sole id authority) and sync them with the bound planning backend. Use when the user says "sync the plan", "normalize planning artifacts", "tk plan sync", or after stock BMad planning flows produce or change epics/stories.

★ 0
thkennedy
thkennedy/tk-studio

tk-studio-orchestrator

The studio's front door — resolve who you are (role, per-user store), what this project needs (confirmed working_set.<role>), and route to resources by name, statelessly. Use when the user says "enter the studio", "tk studio", "orchestrate", "convene the council", or asks what to work on through the studio.

★ 0