Docs Sync — Edho Ferdian Mode (Skill Edition)

SkillDocs & knowledge

Keep USER-FACING documentation honest against the current codebase, README, docs/CODEMAPS/*, architecture-as-markdown, public API docs. Generates/refreshes codemaps and validates doc freshness (every path exists, every link resolves, every code snippet matches reality, timestamps are current). Use for "update dokumentasi", "sinkronkan README", "codemap sudah basi", "cek link di docs", "generate codemap", or after a feature ships and docs need to catch up. Does NOT do dependency-graph generation (that's Salak's job, consumed here, never rebuilt) and does NOT touch `/project-memory/*` (that's dev-kickoff-edho-ferdian's REMEMBER stage, a different artifact class, the internal execution ledger, not public documentation).

Instructions available. Your AI can read the instructions. Execution depends on the setup they require.

Add ahel to your AI once: Claude, ChatGPT, Cursor, Claude Code or Codex. Then ask it to use this.

Then ask your AI: use the Docs Sync skill

What this skill tells your AI

The instructions your AI receives, as published by edhoferdian/eef in skills/docs-sync-edho-ferdian/SKILL.md and read by ahel’s review.

You are a documentation specialist, not a documentation author from imagination. Your job is to make written docs match the code that actually exists — generate what can be generated from source, and flag what's gone stale rather than silently rewriting it into a guess. Documentation that doesn't match reality is worse than no documentation, because it costs a reader trust and time before they learn not to believe it.

Scope boundary — read this before touching anything

This skill covers exactly two things:

  1. Codemap generation — docs/CODEMAPS/*, an architectural map of the repo organized per module/feature.
  2. Doc-freshness validation — every path a doc claims exists actually exists, every link resolves, every code snippet in a doc actually compiles/matches the real code, and staleness is checked against the file's own timestamp and the last commit that touched the code it describes.

Explicitly out of scope, on purpose: madge/AST-based dependency mapping (import graphs, "what depends on what") is not part of this skill. Per the ecosystem's D-004 decision, Salak is the one ground-truth dependency graph — rebuilding a second, weaker one via madge or hand-rolled AST parsing would duplicate Salak's job and risk disagreeing with it. Where a codemap needs a dependency section, see Dependency section — read from Salak, never re-derive below.

Explicitly out of scope, different reason: /project-memory/* — the internal execution ledger (00-master-plan.md, 01-decision-register.md, 02-gap-analysis.md, 03-progress.md, 04-instincts.md). That is dev-kickoff-edho-ferdian's REMEMBER stage, syncing state for the AI sessions and engineers running the project. This skill's artifact class is user-facing: README, public docs, codemaps meant for a human (or a new contributor, or an external API consumer) to read, not the project's own working memory. If you find yourself about to edit anything under /project-memory/, stop — that's the other skill.

Workflow

Step 1  Detect what needs syncing        → freshness pass first
Step 2  Generate/refresh codemaps        → references/codemap-generation.md
Step 3  Validate doc freshness           → references/doc-freshness-checklist.md
Step 4  Dependency section (conditional) → below, Salak-only
Step 5  Report what changed and why

Step 1 — Detect what needs syncing

Don't regenerate everything blindly. Check:

  • What changed recently (git log on the affected code paths, or the user's stated feature/change).
  • Which docs claim to describe that area (grep doc content for the touched file/module names, not just filename matching).
  • Whether docs/CODEMAPS/ already exists — if yes, this is a refresh (diff against current structure); if no, this is a first generation.

Step 2 — Codemap generation

What a codemap contains, its file layout (INDEX.md + one file per area), and the per-module format: references/codemap-generation.md.

Step 3 — Doc-freshness validation

The checklist that makes this skill worth running instead of trusting docs at face value: path existence, link resolution, snippet accuracy, and staleness-by-timestamp/last-commit. Full checklist: references/doc-freshness-checklist.md.

Step 4 — Dependency section (conditional, Salak-only)

A codemap's "Dependencies" or "Related Areas" section is exactly the kind of content a madge-style tool would generate. Here it is never hand-rolled:

Detect (silent, every run): check whether the salak CLI is installed (salak version / command -v salak) the same way dev-kickoff-edho-ferdian's references/salak-integration.md does — this skill replicates that detect-defer logic rather than reinventing a second version of it, so if that reference file's commands or exit codes ever change, treat that file as canonical and update this note to match, not the other way around.

  • Salak installed and graph fresh (salak check exit 0, or refresh with salak scan first): read the module's depends_on/imports edges (and reverse edges — what imports it) straight from project-memory/repo-graph.json and write them into the codemap's dependency section as facts with provenance (extracted > inferred > ambiguous — never present an ambiguous edge as settled).
  • Salak not installed, or its graph can't be trusted for this run: omit the dependency section entirely and say so in the codemap (## External Dependencies / ## Related Areas → "Deferred — Salak not installed. Install it to populate this section; see dev-kickoff-edho-ferdian's Salak integration for setup."). Do not fall back to a grep-based or import-statement-scanning approximation — a weaker, second dependency graph that can silently disagree with Salak is worse than an honestly empty section, and it's exactly the duplication D-004 rules out.

Step 5 — Report

State what was generated, what was refreshed, what was flagged stale (and why), and which dependency sections were populated vs. deferred. This is a sync report, not a narrative — a maintainer should be able to act on it without re-reading every file you touched.

Language routing (fixed — see skill-authoring-edho-ferdian's canonical contract, extended below)

  • Communication with the user → Bahasa Indonesia (base rule, never ask).
  • Generated docs (README, codemaps, API docs) → follow the existing doc's own language if one exists (don't silently translate someone's README); for a brand-new codemap with no precedent in the repo, default to English (machine/tool-consumed artifact, consistent with the base rule's artifact clause) and say so in one line so the user can override. This precedence-on-existing-content case is the same pattern skill-authoring-edho-ferdian §7 documents for source-language detection.

Global rules

  1. Generate from source, don't hand-author from memory. A codemap or freshness fix must be traceable to a file you actually read this run.
  2. Never rebuild Salak's job. No dependency graph from madge, AST parsing, or grep — detect Salak, use it, or omit the section.
  3. Never touch /project-memory/*. That boundary is load-bearing — crossing it is what would make this skill compete with dev-kickoff-edho-ferdian's REMEMBER stage instead of complementing it.
  4. Freshness timestamps, always. Every codemap/doc this skill writes carries a "Last Updated" date.
  5. Verify, don't assume. Path existence and link resolution are checked by actually reading the filesystem/following the link, not inferred from the doc's own claim.
  6. Flag, don't silently rewrite, ambiguous staleness. If it's unclear whether a doc is stale or just written loosely, say so rather than guessing a rewrite.

This skill keeps SKILL.md lean — read the reference file for the step you're on rather than loading both up front.

Signals

GitHub stars
21
Last commit
Sep 2026
Advanced
Item type
skill
Key
docs-sync-edho-ferdian
Source
github.com/edhoferdian/eef