بنقرة واحدة
orchestrate-doc-pipeline
expert: "Sequential pipeline coordination dispatching five editorial
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
القائمة
expert: "Sequential pipeline coordination dispatching five editorial
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Use at the start of every agent session, even if the task seems obvious and you think you already know what to do. Activates when the agent has just opened a repository, does not know what to do, needs to orient itself, or is beginning any new session. Also activates for "where do I start?", "what should I work on?", "what is the current state?". Skipping orientation leads to wasted effort and missed context.
expert: "Structured task execution procedure for individual implementation
expert: "Multi-agent development orchestration with conflict-aware parallel
expert: "Review orchestration with adaptive specialist dispatch,
Use at the start of every agent session, even if the task seems obvious and you think you already know what to do. Activates when the agent has just opened a repository, does not know what to do, needs to orient itself, or is beginning any new session. Also activates for "where do I start?", "what should I work on?", "what is the current state?". Skipping orientation leads to wasted effort and missed context.
expert: "Structured task execution procedure for individual implementation
| name | orchestrate-doc-pipeline |
| description | {"expert":"Sequential pipeline coordination dispatching five editorial stages with changelog collation, advisory human checkpoints, re-entry handling, and completion reporting for documentation refinement","natural":"Run a document through the editorial pipeline — coordinate the five stages, track changes, offer checkpoints, and produce a summary"} |
| triggers | ["run the documentation pipeline","publish a document through the editorial pipeline","editorial pipeline","refine documentation","run document through all editorial stages","coordinate document editing"] |
| roles | ["doc-pipeline-orchestrator"] |
| stage | documenting |
| constraint_level | medium |
Pipeline structure:
Change tracking:
Human interaction:
System integration:
doc(action: register) at pipeline start; refreshing with doc(action: refresh) at pipeline endCopy this checklist and track your progress:
Before starting:
- [ ] The document to be processed is identified and accessible
- [ ] The document is registered with `doc(action: register)` or already exists in the doc system
- [ ] The document's purpose, type, and audience are understood (needed for the Write stage; may already exist)
Per stage (who edits the file?):
- Write, Style, Copyedit → the sub-agent edits the file directly
- Edit, Check → the sub-agent produces a report; the orchestrator applies changes to the file
- [ ] The stage receives the document and the previous stage's changelog (if any)
- [ ] The document file has been modified (by the sub-agent or the orchestrator) — if the stage reported findings but the file is unchanged, something went wrong
- [ ] The stage's output includes a changelog of what was changed and why
- [ ] The changelog is reviewed for boundary violations before passing to the next stage
- [ ] Any stage flags (issues outside scope) are recorded
Checkpoints:
- [ ] Advisory checkpoint offered after Edit (structural review)
- [ ] Advisory checkpoint offered after Copyedit (final review)
Completion:
- [ ] All five changelogs are collated into a completion summary
- [ ] Stage flags are surfaced in the summary
- [ ] The document record is refreshed with `doc(action: refresh)`
- [ ] Any feedback signals (recurring patterns) are noted for future retrospective
If the document is not already registered, call doc(action: register) with the document path, type, and title. This creates the tracking record. IF the document already exists in the doc system, confirm its record is current and note its type and purpose.
Pass the document purpose, type, audience, and source material to the write-docs skill via handoff. If the document already exists (human-written or previously drafted), pass the existing document as the input — the Write stage will use it as the base rather than drafting from scratch. The Write stage edits the file directly — it produces a revised draft (or a first draft) and verification notes. Confirm the output includes a changelog or initial notes describing what was produced.
Pass the draft document and the Write stage's output to the edit-docs skill. Receive: structural edit report with findings classified as structural-blocking or structural-suggestion. The Edit stage produces a report, not a revised file. Review the changelog for boundary violations — the editor should not have rewritten sentences or fixed facts. IF structural-blocking findings exist, the orchestrator applies them to the file before continuing. This is the orchestrator's responsibility — do not pass an unmodified document to the Check stage when there are blocking findings.
Offer the human a chance to review the structural edit. Present the heading skeleton assessment, any structural-blocking findings, and the Edit changelog. IF the human provides feedback → incorporate it and note the feedback in the checkpoint record. IF the human declines or does not respond → continue. Record the checkpoint outcome (reviewed / skipped / feedback incorporated).
Pass the structurally-edited document to the check-docs skill. Receive: QA report with classified findings (hallucination, unverified, stale, vague, inflated, promotional). The Check stage produces a report, not a revised file. The orchestrator applies factual corrections (hallucinations, stale references) to the file. Flag substance issues (vague, inflated) for the Style stage — do not correct those here. Review the changelog for boundary violations — the checker should not have restructured sections or rewritten prose.
Evaluate the Check stage's flags. IF the Check stage flagged severe structural problems (e.g. an entire section based on a hallucinated feature, or sections that need to be removed entirely) → send the document back to the Edit stage (Step 3) and re-run from there. IF issues are minor (a few vague claims, a stale version number) → note them and continue. Re-entry should be rare — reserve it for issues that make later stages pointless.
Pass the fact-checked document and any substance flags from the Check stage to the style-docs skill. The Style stage edits the file directly — it applies all vocabulary and pattern changes, then produces a changelog of what it changed. Review the changelog for boundary violations — the style editor should not have restructured sections or changed factual content. If the file is unchanged but the changelog lists findings, the stage failed to apply its changes — re-dispatch with explicit editing instructions.
Pass the style-edited document to the copyedit-docs skill. The Copyedit stage edits the file directly — it applies all sentence-level changes, then produces a changelog. Review the changelog for boundary violations — the copy editor should not have restructured sections, changed content, or hunted for AI artifacts (that was the Style stage's job). If the file is unchanged but the changelog lists findings, the stage failed to apply its changes — re-dispatch with explicit editing instructions.
Offer the human a final review of the finished document. Present the completion summary with all five changelogs collated, plus any unresolved flags. IF the human provides feedback → apply it and note the feedback in the checkpoint record. IF the human declines or does not respond → continue. Record the checkpoint outcome.
doc(action: refresh).retro tool to surface patterns.At pipeline completion, produce a summary:
## Pipeline Completion Summary
**Document:** {document path}
**Type:** {README | getting-started | manual | reference | design}
**Stages completed:** {5/5 or fewer if re-entry occurred}
### Stage Results
| Stage | Changes | Flags | Key action |
|-------|---------|-------|------------|
| Write | — | {count} | {one-line summary} |
| Edit | {count} | {count} | {one-line summary} |
| Check | {count} findings | {count} | {one-line summary} |
| Style | {count} | {count} | {one-line summary} |
| Copyedit | {count} | {count} | {one-line summary} |
### Human Checkpoint Results
- **Post-Edit:** {reviewed / skipped / feedback incorporated}
- **Post-Copyedit:** {reviewed / skipped / feedback incorporated}
### Re-entry Events
{none, or description of re-entry and reason}
### Unresolved Flags
{issues flagged by stages that were not resolved during the pipeline run}
### Feedback Signals
{recurring patterns to consider for upstream SKILL improvement, if any}
Ran the document through all five stages. The document looks good now.
No issues found.
WHY BAD: No changelogs, no stage summaries, no evidence that any stage did meaningful work. The human cannot tell what changed at each stage or whether boundary violations occurred. This is a rubber stamp of the pipeline, not coordination.
## Pipeline Completion Summary
**Document:** refs/getting-started.md
**Type:** getting-started
**Stages completed:** 5/5
### Stage Results
| Stage | Changes | Flags | Key action |
|-------|---------|-------|------------|
| Write | — | 2 unverified claims | Drafted 8-section guide following getting-started template |
| Edit | 3 structural | 0 | Moved Quick Start before Background; split §4 into two sections |
| Check | 5 findings | 1 for Style | Caught hallucinated --verbose flag; flagged 2 vague claims |
| Style | 12 words, 2 rewrites | 0 | Replaced 8 banned words; rewrote 2 fingerprint-cluster passages |
| Copyedit | 7 sentences | 0 | Fixed 4 passive→active; split 2 long sentences; standardised contractions |
### Human Checkpoint Results
- **Post-Edit:** Reviewed. Approved section reorder.
- **Post-Copyedit:** Skipped (human did not respond within timeout).
### Re-entry Events
None.
### Unresolved Flags
- Check flagged "performance improvement of up to 10x" as unverified — no benchmark source found. Left for author to confirm or remove.
### Feedback Signals
- Third document in a row where Style caught "leverage" and "utilize" — consider adding to Write SKILL anti-patterns.
WHY GOOD: Every stage's contribution is visible. The human can see what changed and why at each stage. Boundary violations are implicitly absent (each stage's changes match its scope). Checkpoints were offered. The completion summary collates all five changelogs into a reviewable record. A feedback signal was identified for upstream improvement.