Agent-Skills.md

Agent Skills: Checkpoint Brief

Project completed stage Artifacts into a concise human-review block in `DECISIONS.md`; use before a Workflow checkpoint, not for Workflow routing or approval.

UncategorizedID: willoscar/research-units-pipeline-skills/checkpoint-brief
51039

Install this agent skill to your local

pnpm dlx add-skill https://github.com/WILLOSCAR/research-units-pipeline-skills/tree/HEAD/.codex/skills/checkpoint-brief

Skill Files

Browse the full folder contents for checkpoint-brief.

Download Skill

Loading file tree…

.codex/skills/checkpoint-brief/SKILL.md

Skill Metadata

Name
checkpoint-brief
Description
Project completed stage Artifacts into a concise human-review block in `DECISIONS.md`; use before a Workflow checkpoint, not for Workflow routing or approval.

Checkpoint Brief

Prepare the review surface between machine-produced Artifacts and a human checkpoint. This Skill summarizes; it does not choose a Workflow, approve a checkpoint, or modify semantic Artifacts.

Inputs

  • the active PIPELINE.lock.md
  • the stage Artifacts named by the current Unit
  • the existing DECISIONS.md

Output

  • one current checkpoint block in DECISIONS.md

Steps

  1. Read the locked Workflow and current checkpoint.
  2. Inspect only the declared stage Artifacts.
  3. Summarize scope, structure, coverage, and unresolved choices without adding new research claims.
  4. Upsert the checkpoint block while preserving the approval checklist and prior Decision history.
  5. Stop before approval. human-checkpoint owns the approval Decision.

Acceptance

  • DECISIONS.md names the current checkpoint and the Artifacts reviewed.
  • unresolved choices are explicit and bounded.
  • no approval is inferred or written.
  • no Pipeline lock, Unit state, or semantic Artifact is changed.

Script

uv run python .codex/skills/checkpoint-brief/scripts/run.py \
  --workspace workspaces/<name> \
  --checkpoint C2 \
  --inputs 'outline/taxonomy.yml;outline/outline.yml'

The Pipeline adapter supplies the current Unit's declared inputs. Manual invocations must do the same; an empty input list is reported as unreviewable rather than inferred from unrelated Workspace files.