From superRA
Use whenever creating a PLAN.md / RESULTS.md from scratch, maturing RESULTS.md into its permanent record at INTEGRATE, or when you need the full editing discipline for a task-block-structured handoff document. Carries the four document principles, the inline-edit rule, the stale-content checklist, and pointers to full PLAN.md / RESULTS.md anatomy templates. Usable standalone by a single author with no subagents — the author plays all roles and reads this skill directly. Doc-creation call sites: `superplan` Phase 2 (new plan + RESULTS.md skeleton) and `superintegrate` Step 3 doc-writer (Stage 2 maturation).
How this skill is triggered — by the user, by Claude, or both
Slash command
/superRA:handoff-docThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Handoff docs (`PLAN.md`, `RESULTS.md`, and similarly-structured task-block docs) are the persistent state of a project — multiple agents and sessions read and write them. References:
Handoff docs (PLAN.md, RESULTS.md, and similarly-structured task-block docs) are the persistent state of a project — multiple agents and sessions read and write them. References:
references/plan-anatomy.md — full PLAN.md template and the User Decisions Log spec.references/results-anatomy.md — full RESULTS.md template and the two-stage RESULTS.md lifecycle (Stage 1 dev log → Stage 2 permanent record).Subagent review-loop mechanics live in agents/implementer.md and agents/reviewer.md.
Latest state only, no history. Handoff docs reflect current intent and current findings. They are not changelogs — git owns history. No "Previously...", no strikethroughs, no "Update:" blocks, no stacked review rounds.
Live and committed. Every edit is an inline replacement, committed atomically with the work it belongs to. Stale steps, stale review notes, and superseded discovery notes are removed, not struck through. The doc at any point reads as a single coherent current-state description.
Task-block structure. PLAN.md is a header plus a sequence of task blocks; each task block has a fixed anatomy. RESULTS.md mirrors the task structure. The full templates live in references/plan-anatomy.md and references/results-anatomy.md.
The doc is the record. Status reports are pointers, not substitutes. Any material finding, result, methodology change, caveat, or decision MUST be written into PLAN.md or RESULTS.md before it is communicated in a status report or chat message. If a result exists only in chat, it does not exist — it will be lost at the next session boundary, cache eviction, or context compaction.
Rule of thumb: before typing a finding into a status report, ask "is this written in PLAN.md or RESULTS.md yet?" If not, write it in the doc first and commit, then point at it in the report.
Every edit replaces stale content in place. Never append, never strike through, never use "Update:" / "Revised:" / "Previously..." framing. History belongs in the git commit message.
Common stale content to replace in place:
Researcher answers to AskUserQuestion / plain-text pauses are written into PLAN.md before the agent acts on them, committed atomically with the work they unblock. The decision is not resolved until it is in the record. Full contract: using-superra/references/main-agent.md.
Full spec — where task-scoped vs project-level decisions land, the three-line blockquote format, the hook reminder, and what does NOT count as a decision — lives in references/plan-anatomy.md §User Decisions Log.
## Project Conventions — populated at superplan Phase 3; anatomy in references/plan-anatomy.md §Project Conventions.
Figure embedding and code-file citations — discipline in report-in-markdown/references/rich-content.md; Stage 2 materialization in report-in-markdown/references/final-form.md.
Guides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.
npx claudepluginhub fuzhiyu/superra --plugin superRA