AgentSkillsCN

03 Plan Builder

03 计划构建者

SKILL.md
--- frontmatter
tags: [skill, codex-exec]
skill_id: "03_plan_builder"

Skill 03 — Plan Builder

Purpose: Produce an implementation plan that is grounded in research evidence and acceptance criteria.

Protocols:

  • [[docs/PROTOCOLS/VAULT_PROBE]]
  • [[docs/PROTOCOLS/EVIDENCE]]
  • [[docs/PROTOCOLS/RUN_LOGGING]]
  • [[docs/PROTOCOLS/PATCH_GATE]]
  • [[docs/PROTOCOLS/CONTEXT_BUDGET]]

Tools:

  • [[tools/repl/README]] — patch-gated edits (generate + check + apply)

Inputs (from Orchestrator)

  • docs/tasks/<task_slug>.md (required)
  • docs/research/<task_slug>.md (required)

Outputs (must write)

  • docs/plans/<task_slug>.md
  • artifacts/plan_steps.json
  • [[docs/skills/03_plan_builder/SUMMARY]]

Run log (mandatory)

Create: artifacts/_runs/<timestamp>_03_plan_builder/


Steps

  1. Slice-read the task + research notes (do not read entire docs).
  2. Draft a plan with:
    • milestones
    • ordered steps
    • file-level targets (paths)
    • verification steps (tests, probes)
  3. Identify unknowns / dependencies and propose how to de-risk them.
  4. Write docs/plans/<task_slug>.md.
  5. Write artifacts/plan_steps.json with an ordered array:
    • {step_id, description, files, verification, risk}
  6. Write SUMMARY.md with:
    • plan link
    • next skill recommendation (usually Change Executor)

Definition of Done

  • Plan exists and references acceptance criteria
  • Plan includes verification steps (tests/probes)
  • Plan JSON exists and lists concrete file paths or search targets
  • Summary exists