$ ~/registry/skill/openclaw-clawdinator-workspace-skills-distill-pr-intent-orchestrator

SKILL

distill-pr-intent-orchestrator

Name: distill-pr-intent-orchestrator
Author: openclaw

Distill intent from multiple pull requests and save structured outputs for analysis.

Stars

★ 157

Source

GitHub

Updated

2026-06-27

// securityLow risk

Prompt only, no code
Open source
Community-trusted· 157

Running security audit…

Credentials
Network egress
Code execution
Data access
Supply chain

// install

Copy the install command and let the AI configure it · recommended for beginners

Please install the "distill-pr-intent-orchestrator" skill from askskill:
1. Download https://raw.githubusercontent.com/openclaw/clawdinators/main/clawdinator/workspace/skills/distill-pr-intent-orchestrator/SKILL.md
2. Save it as ~/.claude/skills/distill-pr-intent-orchestrator/SKILL.md
3. Reload skills and tell me it's ready

// download

Download SKILL.mdInstall manifest ↗

// examples

Batch PR intent distillation

Prompt

Please distill the change intent, key impact, and risks for each PR in this list, then save the results and sidecar metadata to /memory/pr-intent.

Expected

Structured intent summaries, impact notes, risk analysis for each PR, and persisted metadata files.

Generate summaries for release review

Prompt

Analyze all PRs related to this release, extract each PR's business purpose, technical change type, and potential release impact, then save the output to /memory/pr-intent.

Expected

A release-review-ready PR intent digest with traceable sidecar metadata.

Capture code review knowledge

Prompt

Run intent distillation on the past week's PR set, summarize common change patterns, goals, and review considerations, and persist all results to /memory/pr-intent.

Expected

A reusable knowledge base of PR intents for future review, search, and team retrospectives.

// docs

Distill PR Intent — Orchestrator

Goal

Run distill-pr-intent across a selected PR set, and persist:

primary memo text (.txt)
sidecar metadata (.meta.json)

All outputs go to shared memory on EFS, and are automatically published to the public S3 bucket by the host timer.

Inputs

repo: fixed openclaw/openclaw
selection: one of:
- last N (e.g. last 500)
- range <start..end>
- since <YYYY-MM-DD>
skill_version: string (default v1)

Output paths (persisted)

/memory/pr-intent/<skill_version>/openclaw-openclaw/<PR>.txt
/memory/pr-intent/<skill_version>/openclaw-openclaw/<PR>.meta.json

Determinism rule

NEVER put timing/telemetry/debug into the primary .txt memo.
All telemetry goes into .meta.json.

Workflow

Determine PR list.

Examples:

# last N merged PR numbers
gh pr list -R openclaw/openclaw --state merged --limit 200 --json number --jq '.[].number'

# range
seq <start> <end>

For each PR (sequential by default):

Run distill-pr-intent <PR>.
Write memo to /memory/pr-intent/.../<PR>.txt.
Write metadata to /memory/pr-intent/.../<PR>.meta.json.

Writing files (memory locking)

Use the locking helpers (required on EFS):

# memo
printf '%s\n' "$MEMO" | memory-write "/memory/pr-intent/<skill_version>/openclaw-openclaw/<PR>.txt"

# meta
printf '%s\n' "$META_JSON" | memory-write "/memory/pr-intent/<skill_version>/openclaw-openclaw/<PR>.meta.json"

Suggested metadata schema

{
  "pr": 17392,
  "repo": "openclaw/openclaw",
  "state": "merged",
  "skill_version": "v1",
  "generated_at": "2026-02-15T19:00:00Z",
  "patch_mode": "PATCH_OK",
  "patch_bytes": 12345,
  "model": "openai/gpt-5.2-codex",
  "error": null
}