SiTaggart Read PStack runtime guidance before this workflow. It defines tool mappings and overrides upstream host assumptions.
Reflect
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
Use the host history workflow in RUNTIME.md. Identify this project and the exact task before reading messages. If the transcript is unavailable, use a labeled session digest and report the missing evidence.
2. Spawn three reviewers in parallel
One message, three native subagent calls, a general-purpose native subagent, explicit model: on each, a read-only brief. Use the tools exposed to the child; the parent applies edits.
| Lens | model |
Prompt template |
|---|---|---|
| Judgment | the reflect judgment, divergent, synthesizer role resolved per RUNTIME.md |
references/judgment-reviewer.md |
| Tooling | the reflect tooling role resolved per RUNTIME.md |
references/tooling-reviewer.md |
| Divergent | the reflect judgment, divergent, synthesizer role resolved per RUNTIME.md |
references/divergent-reviewer.md |
Pass each template verbatim, substituting the transcript path or digest where marked. Reviewers return findings in their completion messages; background spawning returns a handle; a foreground call may return the result directly. Follow the native tool’s documented completion contract.
3. Synthesize
One native subagent call, a general-purpose native subagent, using the reflect judgment, divergent, synthesizer role resolved per RUNTIME.md, a read-only brief. Use available tools to spot-check citations. 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.
List backlog items locally. File tickets only if the user has authorized publication to that tracker.
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): follow the selected adapter’s skill-authoring workflow and its draft / test / iterate loop.
tune description: <skill path>(the skill exists but didn't trigger when it should have): hand toskill-creatorand check its description and invocation policy against realistic trigger examples.new skill via skill-creator: <kebab-name>: hand creation toskill-creator. 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.