Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/ghostlygawd/codeweb/narrategit clone --depth 1 https://github.com/GhostlyGawd/codewebWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00051 | $0.00461 |
| Opus 5 | $0.00026 | $0.00230 |
| Sonnet 5 | $0.00010 | $0.00092 |
| Haiku 4.5 | $0.00005 | $0.00046 |
Grade A, and why
codeweb-narrate scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
/codeweb-narrate
The fast path's map answers "how big" but never "what for" — every node summary is empty and
every domain summary is a templated count. This command restores the English meaning as a
sidecar with provenance (AI-IDEAS.md Idea 3): graph.json stays byte-identical; delete the
sidecar and you have today.
Hard rules (the fence)
- Write ONLY
.codeweb/narration.json. Never editgraph.json, never any source file. - One sentence per entry. What the domain/symbol is FOR — not how big (the map already says).
- Stamp it against the graph you read, so staleness is visible and stale narration silently drops out of every surface (same discipline as the other sidecars).
- Rendered surfaces label every sentence agent-written — narration annotates the map; it never is the map.
Steps
- Read the map:
codeweb_brieffor the domain list;node scripts/reading-order.mjs <graph>for what to read first; read the top files per domain (bounded — this is a sentence, not a review). - Stat the graph for the stamp:
mtimeMs+sizeof.codeweb/graph.json. - Write
.codeweb/narration.json:
{
"version": 1,
"stamp": { "graphMtimeMs": <graph mtimeMs>, "graphSize": <graph bytes> },
"domains": { "<domain name>": "<one sentence — what it does>" },
"symbols": { "<node id>": "<one sentence>" }
}
Cover every product domain; cover the top ~20 load-bearing symbols (the brief's list). Re-run after a re-map to refresh the stamp.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 40 lines · 51 tokens per session scan A 1968703f2b64
codeweb-narrate is a command published in the GitHub repository GhostlyGawd/codeweb (0 stars, last pushed yesterday), licensed MIT. It adds 51 tokens to every session and 461 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
map
Generate MAP.md — a relational map of every file, function, signature, and call in the repo.
sanity
Cross-check a dekko callers/uses result against a targeted grep before trusting a low/zero count.
doctor
Diagnose dekko environment/install issues (PATH shadowing, map freshness, MCP registration, plugin, hooks, CLAUDE.md policy block).
review
Review the implementation of the plan.
compliance
이전 보안 진단 보고서의 Finding들이 패치되었는지 확인하고, 변경된 코드에서 신규 취약점을 탐색합니다.
va
Vulnerability Assessment — 8차원 아키텍처 진단 + Self-Verify + Evidence Verification.