Skip to content

jscraik/Agent-Skills / Infrastructure/references/deferred-skill-context/skill-factory-skillify

skillify

Verified

Capture a completed Codex workflow as a reusable SKILL.md package by analyzing session context plus optional session-collector evidence, interviewing the user with structured prompts, and writing a validated skill artifact. Use when the user asks to skillify or operationalize a repeatable process.

references/session-collector-intake.md

View on GitHub
# Session Collector Intake

Read when: `skillify` is converting a completed conversation, run, or repeated agent workflow into a reusable skill package.

Use this intake to turn session evidence into stable skill behavior without copying raw transcripts into public skill files.

## Intake Boundary

- Use `skillify` to convert a selected, repeatable workflow into a skill.
- Use `skill-refactor` first when the user asks what happened across many sessions, which sessions matter, or whether a workflow should be merged, pruned, improved, or skillified.
- Do not make `skillify` responsible for broad session-history inventory.

## Evidence To Consume

Prefer bounded extracted evidence:

- `~/.agents/session-collector` bundle files when available, especially `skillify-candidates.json`, `index.json`, and `redaction-report.json`
- user goal and repeated trigger language
- assistant procedure that actually worked
- commands or tools that are essential to the workflow
- validation gates that proved the result
- blockers and failure boundaries that future agents must respect
- examples of should-trigger and should-not-trigger user wording

Avoid:

- raw full transcripts
- private file trees in examples
- secrets, tokens, account identifiers, or personal content
- one-off debugging chatter that is not part of the reusable process

## Preferred Collector Run

When the user points to prior sessions but has not supplied a compact handoff, run or request a local session-collector bundle before drafting:

```bash
cd ~/.agents/session-collector
UV_CACHE_DIR=/tmp/session-collector-uv-cache \
uv run --python 3.12 python main.py \
  --days 30 \
  --max-sessions 200 \
  --output /tmp/skillify-session-evidence.json \
  --bundle-dir /tmp/skillify-session-evidence
```

Use `/tmp/skillify-session-evidence/skillify-candidates.json` as the primary workflow-candidate input, `/tmp/skillify-session-evidence/index.json` for session labels, and `/tmp/skillify-session-evidence/redaction-report.json` to confirm sensitive content stayed out of reusable skill artifacts. Do not copy the aggregate raw transcript payload into `SKILL.md`.

## Extraction Handoff

When another skill or collector provides session evidence, ask for or create a compact handoff:

```text
schema_version: 1
source: session-collector
collector_bundle: <local bundle path>
workflow_name: <plain-language workflow>
repeatability: <high|medium|low>
trigger_phrases:
  - <phrase>
required_inputs:
  - <input>
procedure:
  - <step>
validation:
  - <gate>
failure_boundaries:
  - <condition that should stop or redirect>
confidence: <high|medium|low>
limitations:
  - <missing or weak evidence>
redactions:
  - <what was removed>
```

Proceed only when repeatability is at least medium and the reusable behavior is clearer than the incidental details of the session.

## Skill Drafting Rules

- Generalize from the extracted workflow, not from a single timestamp, branch name, or machine path.
- Preserve validation and safety gates that prevented failures.
- Move long examples or nuanced caveats into references and signpost them from `SKILL.md`.
- Add eval cases for happy, edge, failure, and pressure behavior based on the extracted evidence.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall