alexanderop Reflect
Before delegating, read the harness instructions and the active harness file. Map agent names, tool arguments, and model roles there.
Mine the current conversation for durable learnings, then route them into skill edits.
When to invoke
Invoke when the user says "reflect" or "/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.
Process
1. Locate the active transcript
The parent finds its own transcript file before fanning out. Read the
"Transcripts" section of harness/<detected>.md for where they live and in what
form — it differs on all four harnesses, and on Codex and Copilot CLI there is no
per-workspace directory to list. Scope to the active workspace only; globbing
across all workspaces reads private chats from unrelated projects.
Where the harness does give you a per-workspace directory (Cursor, Claude Code), newest-first by real modification time:
ls -t <transcripts>/*.jsonl <transcripts>/*/*.jsonl <transcripts>/*/subagents/*.jsonl 2>/dev/null | head -10
On Codex, bound by date under ~/.codex/sessions/<YYYY>/<MM>/<DD>/ and match cwd.
On Copilot CLI, query ~/.copilot/session-store.db by cwd for the session ids,
then read each ~/.copilot/session-state/<id>/events.jsonl.
Three transcript layouts: legacy flat (<id>.jsonl), current nested (<id>/<id>.jsonl), and subagent (<parent>/subagents/<child>.jsonl).
For each candidate, read the first JSONL line and check that message.content[0].text contains the conversation's opening user prompt. Take the matching path. If no path resolves, write a tight digest of the session and pass that instead.
2. Spawn three reviewers in parallel
One message, three subagent calls, the harness's general-purpose agent type, explicit model: on each, agent mode (readonly: false). Reviewers need MCP access for context lookups (tickets, chat threads, observability traces referenced in the transcript). Readonly strips MCPs.
| Lens | model |
Prompt template |
|---|---|---|
| Judgment | your configured reflect-judgment model (harness default) | references/judgment-reviewer.md |
| Tooling | your configured reflect-tooling model (harness default) | references/tooling-reviewer.md |
| Divergent | your configured reflect-judgment model (harness default) | references/divergent-reviewer.md |
Pass each template verbatim, substituting the transcript path or digest where marked. Reviewers return findings in the Task response body.
3. Synthesize
One subagent call, the harness's general-purpose agent type, using your configured reflect-judgment model (harness default), agent mode (readonly: false). The synthesizer's quality check includes spot-verifying citations, which can require MCP access. Readonly strips MCPs. Use references/synthesizer.md verbatim, with each reviewer's full output inlined where marked. The synthesizer returns a structured Accepted / Rejected / Backlog list.
4. Structural enforcement check
Sanity-check the synthesizer's Accepted list. For any item that would be enforced more reliably by a lint rule, script, metadata flag, or runtime check, move it from Accepted to Backlog. See the encode-lessons-in-structure principle skill.
5. Apply
Before applying any Accepted edit, present the synthesizer's full Accepted/Rejected/Backlog output to the user and wait for explicit approval. The user picks which subset to apply and may redirect routings. Skill changes affect every future agent in the org. Do not auto-apply.
Backlog items file to whatever devex / backlog tracker your team uses automatically. Only the Accepted list waits for approval.
For each approved Accepted item, follow the Routing field exactly:
- Trivial existing-skill edit (a one-line bullet, a tightened sentence, a stale fact corrected): parent does directly.
- Substantive existing-skill edit (a new section, a new pattern table, more than ~10 lines): hand to the harness's skill-authoring tool (
harness/built-ins.md) and run its draft / test / iterate loop. tune description: <skill path>(the skill exists but didn't trigger when it should have): hand tocreate-skilland run its description-optimization loop.new skill via create-skill: <kebab-name>: hand creation tocreate-skill. Do not invent the shape ad hoc.
If your environment ships a SKILL.md validator, run it on every touched skill before declaring done. Skip this step if it doesn't.
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.