sync-process-docslisted

Sync docs/PROCESS.md with the actual implementation. Reads all scripts, hooks, prompts, and config files, then updates docs/PROCESS.md to reflect reality. Use when scripts have changed, new features were added, or to verify documentation accuracy. Trigger: "sync process docs", "update process", "process docs aktualisieren"
lx-0/llm-wiki · ★ 1 · AI & Automation · score 70

Install: claude install-skill lx-0/llm-wiki

# Sync Process Documentation ## What This Does Reads the actual implementation (scripts, hooks, prompts, config) and updates `docs/PROCESS.md` to match. Ensures documentation never drifts from code. Run this skill from the engine repo root (the directory that contains `scripts/`, `hooks/`, and `docs/PROCESS.md`). ## Steps ### Step 1: Inventory all scripts ```bash ls scripts/*.py hooks/*.py ``` For each script, extract: - Argparse arguments (flags, defaults) - Key constants (thresholds, timeouts, models) - Import dependencies (especially shared modules: `ollama_client`, `flush_pipeline`, `_transcript`) - Main flow (what it does, in what order) Pay special attention to the single-source modules — they own cross-cutting concerns and must be referenced correctly in the docs: - `scripts/ollama_client.py` — every Ollama call (chat / chat_schema / chat_vision) - `scripts/flush_pipeline.py` — staged-flush state machine - `hooks/_transcript.py` — shared transcript walker for both hooks ### Step 2: Read current docs/PROCESS.md Read the full `docs/PROCESS.md` and identify each documented process section. ### Step 3: Diff documentation vs implementation For each process in `docs/PROCESS.md`, verify: - Scripts mentioned exist - CLI flags match argparse definitions - Constants match (MIN_TURNS, MAX_RETRIES, COMPILE_AFTER_HOUR, etc.) - Flow description matches actual code logic - Edge cases listed are actually handled in code - Routing tables match actual if/elif chains - LLM c