audience: human-expert
status: stable
last-verified-against-commit: HEAD
generated-by: hybrid
docforge-sync
Sub-skill of the docforge pack. Sensor + diagnostician + verifier for doc-code state.
Drift is defect, not feeling.
TL;DR
This skill quantifies documentation drift against code (DDS-1 metric), verifies docs via doctest dispatch + Vale + lychee + frontmatter schema, derives reference markdown from schemas (Pydantic / OpenAPI / Proto / GraphQL), and detects translation lag + diagram alt-text gaps.
The skill body lives in SKILL.md. Read that first.
Directory layout
docforge-sync/
├── SKILL.md # primary skill body (<=500 lines, voice-DNA respected)
├── README.md # this file
├── references/ # progressive disclosure — load on demand
│ ├── drift-classes.md # D1-D5 with detection algorithms
│ ├── dds1-formula.md # DDS-1 components, weights, thresholds, calibration
│ ├── doctest-runners.md # per-language dispatch table
│ ├── vale-rules.md # custom rules + regex
│ ├── frontmatter-schema.md # contract reference
│ ├── traceability-matrix.md # feature x ADR x test x doc algorithm
│ └── ci-templates.md # GitHub Actions + GitLab CI + pre-commit guidance
├── schema/ # canonical JSON schemas
│ ├── frontmatter.schema.json # imported by master skill
│ ├── docforge.config.schema.json
│ └── repo-inventory.schema.json
├── vale-styles/ # Vale custom rule files
│ ├── docforge-A1-hedge.yml # anti-pattern A1 detection
│ ├── docforge-A5-anonymous.yml # anti-pattern A5 detection
│ ├── blameless.yml # postmortem-scoped blame language
│ └── redaction.yml # secret/internal-URL leak safety net
├── scripts/ # functional stubs with full architecture
│ ├── scan.py # build repo-inventory.yaml
│ ├── check.py # compute DDS-1
│ ├── verify.py # orchestrated lifecycle gate
│ ├── parsers/
│ │ └── tree_sitter_universal.py # 305+ language substrate
│ └── lib/
│ ├── exit_codes.py # semantic exit code constants
│ ├── frontmatter.py # parser + validator
│ └── git_helpers.py # last-touched, churn, paired-edit
├── ci-templates/
│ └── github-actions.yml.tmpl # ready-to-copy workflow
└── tests/ # fixtures + smoke tests (to be expanded)
Status of this build
This is the v0.1 skill architecture forged in P4 by skill-architect. Scripts are
functional stubs — full method signatures, complete docstrings, architecture decisions
codified, but tree-sitter / Vale / lychee integrations are marked # TODO. Implementation
follows in subsequent PRs; the contracts (schemas, CLI surface, exit codes, output formats)
are stable.
What is complete and authoritative in this build:
SKILL.md body — Anthropic-grade, voice-DNA respected, trigger keywords explicit.
- All
references/*.md — algorithm specs, thresholds, calibration notes.
- All
schema/*.json — canonical contracts (master skill imports frontmatter.schema.json).
- All
vale-styles/*.yml — production-ready Vale rules.
scripts/ — function signatures, dataclasses, exit-code semantics, CLI argparsers — all in place.
ci-templates/github-actions.yml.tmpl — directly usable.
Sibling and parent skills
- Parent (master):
docforge — orchestrator. Routes to this skill for drift + verify + derive operations.
- Sibling:
docforge-llms — projection canonical -> per-page .md / llms.txt / AGENTS.md / CLAUDE.md / MCP scaffold.
Provenance
Forged in D:/lozano-system/workspace/docforge-build/P4-skills-forge/ from:
P3-athena-framework/prd.md v1.0 + prd-addendum-v1.1.md
P3-athena-framework/identity-persona.md (5-layer architecture)
P3-athena-framework/values-ranking.md (E1>E5>E2>E4>E3 + tiebreaks)
P3-athena-framework/voice-dna.md (7 signature phrases + 7 anti-phrases)
P2-research/C-drift-sot.md (full technical architecture)
P2-research/D-philosophy.md (anti-patterns A1-A5)
Auto-validation summary
| Test | Result |
|---|
| (a) Trigger keywords explicit in description | PASS — drift, doc-code sync, DDS-1, doctest, Vale, lychee, AST, tree-sitter all present |
(b) SKILL.md <= 500 lines | PASS — well under limit |
| (c) Voice DNA respected (no hedge words, signature phrases used) | PASS — SP3 ("Drift is defect, not feeling.") in TL;DR + check.py output; confidence labels in references |
| (d) Frontmatter contract spec present | PASS — references/frontmatter-schema.md + schema/frontmatter.schema.json |
| (e) >=1 concrete example per capability | PASS — every capability section in SKILL.md has a CLI invocation + output sample |
| (f) Self-reference test | PASS — SKILL.md has its own frontmatter + zero hedge soup + zero best-practice anonymous + last-verified set |

<sub>Built with MyClaude Studio Engine</sub>