learn-reflect
Mine this session for durable learnings with three parallel reviewers, and route each accepted one to a concrete edit on an existing skill. Use when the user says reflect.
- License 许可
- MIT
- 最近更新
- 2026-11-05
- Stars
- ★ 1
- Forks
- 0
01 前言 Front Matter
- name
- learn-reflect
- description
- Mine this session for durable learnings with three parallel reviewers, and route each accepted one to a concrete edit on an existing skill. Use when the user says reflect.
- disable-model-invocation
- true
02 资源文件 6
- 参考文档agents/divergent-reviewer.md3.9 KB
- 参考文档agents/judgment-reviewer.md3.3 KB
- 参考文档agents/openai.yaml164 B
- 参考文档agents/synthesizer.md4.4 KB
- 参考文档agents/tooling-reviewer.md4.6 KB
- 参考文档CREDITS.md1.1 KB
03 SKILL.md 正文
Reflect
Mine the current conversation for durable learnings, then route them into skill edits.
Invoke when the user says "learn-reflect" or "/learn-reflect". Skip when the conversation is trivial, off-topic, or already covered by an existing skill the parent followed correctly. One-offs are not learnings.
1. Locate the active transcript
Before fanning out, find this conversation's transcript in the harness transcript store, in this order: first check consumer-repo config under docs/agents/ for a transcript store path if that config is present, then ask the user once for the transcript store path, then detect a readable store only through generic user-set signals such as an environment variable, and finally fall back to a tight digest. Read only this workspace's store: the current project's session files.
Match the file whose opening user prompt is this conversation's first message. Never glob across sibling buckets or other projects: that crosses workspace boundaries and reads private chats from unrelated projects. If no file matches, or no readable store was found, pass a tight digest of the session instead (what was tried, what failed, what got corrected, what was decided and why). A digest plus the PR diff or run log you already have beats a guessed transcript path.
2. Spawn three reviewers in parallel
One message, one batch of three reviewer spawns through the harness subagent tool. Reviewers need their normal tool access to look up the tickets, chat threads, and observability traces the transcript references; a stripped tool set weakens them, so do not run them read-only.
Models follow the repo model map: learn-reflect-reviewer for the three reviewers (missing file or row means the parent model). No harness-specific agent type and no model pin appear in the spawn; copy the role id through unchanged. With a single model, run two passes (find, then challenge the findings). Reviewers never edit files; they return findings in their response body.
| Lens | Prompt template |
|---|---|
| Judgment | agents/judgment-reviewer.md |
| Tooling | agents/tooling-reviewer.md |
| Divergent | agents/divergent-reviewer.md |
Pass each template verbatim, substituting the transcript path or digest where marked.
3. Synthesize
One spawn on the learn-reflect-synthesizer role (missing file or row means the parent model), same tool access as the reviewers: it spot-verifies citations through lookup tools. Pass agents/synthesizer.md verbatim, with each reviewer's full output inlined where marked. It returns an Accepted / Rejected / Backlog list.
4. Structural enforcement check
Move any Accepted item that a lint rule, script, metadata flag, or runtime check would enforce more reliably to Backlog. See the principles skill file encode-lessons-in-structure.
5. Apply
Present the synthesizer's full Accepted / Rejected / Backlog output and wait for explicit approval before applying any Accepted edit. The user picks the subset and may redirect routings. Skill changes affect every future agent in the org. Do not auto-apply.
File each Backlog item to your team's devex or backlog tracker without waiting. Only the Accepted list waits for approval.
Never edit a skill inside a vendored or cached copy. Before approval, show each edit's destination: this repo's skills/<bucket>/<skill>/ path for repo skills, or the target repo's skill path for skills living elsewhere. After approval, edit that path directly.
Follow each approved row's Routing exactly:
- Trivial existing-skill edit (a one-line bullet, a tightened sentence, a stale fact corrected): the parent does it directly.
- Substantive existing-skill edit (a new section, a new pattern table, more than ~10 lines): draft it through the whips
authoring-a-skillplaybook and its validate loop. tune description: <skill path>(the skill exists but did not trigger when it should have): retune its description through the same playbook.new skill via create-skill: <kebab-name>: hand creation to the same playbook. Do not invent the shape ad hoc.
Each accepted learning becomes a small PR to its target skill (one skill per PR, small ordered commits) closed through the whips opening-a-pr playbook, or a listed diff when the user wants to land it by hand. If this repo ships a SKILL.md validator, run it on every touched skill before declaring done.
6. Summarize for the user
Short list, no preamble:
- Edits applied:
<skill path>. What changed, one line each. - New skills created:
<skill path>. One line each (rare). - Backlog filed to the devex tracker:
<issue title>(<tags>). One line each. - Dropped: one line per rejected finding + reason from the synthesizer.
04 同仓库更多 Skills
whips
Route any non-trivial task through one entry: match a playbook, drive HIT skills plus subagents, open a reviewed ready PR with ship-pr-shaped evidence.
verify-upkeep
Periodic upkeep for a project's verification skill: source wave plus live pass per feature, at most one PR of proven corrections. Use to audit the verify skill.
verify-statechart
Check a TLA+ model of a system that moves between states, such as a product flow, a protocol, or a service, and draw it as a Mermaid statechart. The report gives the shortest path to each state that breaks a rule. Use to show a bug as a path through states.
verify-new
Generate a project-local skill that drives the app like a user (launch, doctor, drive, evidence, feature map), proven once. Use when a project has no scripted proof.
ship-triage
Move issues and external PRs through a state machine of triage roles, categorise, verify, interview if needed, and write agent-ready briefs.
ship-runner
Run ready-for-agent issue queues on any schedule through /whips intake, one fresh worktree each, landing per risk tier, with the agent and model you pick at setup.