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/applygit 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 | $0.00052 | $0.00751 |
| Opus 5 | $0.00026 | $0.00376 |
| Sonnet 5 | $0.00010 | $0.00150 |
| Haiku 4.5 | $0.00005 | $0.00075 |
Grade A, and why
codeweb-apply 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.
How it starts
The opening of the file, as written. The whole thing — 50 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/codeweb-apply
Execute the gated worklist that /codeweb ends at. Today the advisory stops at a plan
("the agent (+ the gate) executes each step"); this command is that execution — ready-tier
merges only, every accept/reject decided by the deterministic gate, never by you.
Hard rules (AI-IDEAS.md Idea 1 — the fence)
- Ready tier only. Never touch
blockedorreviewtier items — blocked means the gate already rejected the naive merge; review means a human judgement the user has not delegated. - The gate owns the verdict. You simulate, edit, and run tests;
codeweb_diff(exit 1 on a new cycle / new duplication / lost callers) and the test subset decide. You never overrule them. - Revert on any failure. A red diff-gate or failing test subset →
git checkoutthe touched files (orgit revertthe step commit), record the step as blocked with the gate's reason, and CONTINUE to the next step (campaign ordering keeps later steps valid). - Stop conditions: first unexpected error class (not a per-step gate failure), a dirty working tree you did not create, or the user interrupting. Never force-push, never touch CI.
- codemod --write is CLI-only by design — the MCP surface stays read-only; the user's agent writes, codeweb never does.
The loop
- Preconditions: clean
git status; a fresh map (codeweb_mapornode scripts/run.mjs .). - Get the plan:
codeweb_campaign(ornode scripts/optimize.mjs <graph> --json) — take the ready tier in listed order. - Per step:
a. Pre-flight:
codeweb_simulatewith the step's merge (merge,into) — skip the step as blocked ifprojected.okis false (record why). b. Snapshot: copygraph.jsonto abefore.jsonfor this step. c. Apply:node scripts/codemod.mjs <graph> --merge <ids> --into <id> --writefor mechanical merges; hand-edit only when codemod declines, keeping the edit minimal. d. Refresh:codeweb_refresh(ornode scripts/refresh.mjs <graph>). e. Gate:codeweb_diffbefore vs after — exit 1 → revert (rule above). f. Tests:codeweb_testsfor the surviving symbol → run exactly that subset. Red → revert. g. Commit: one commit per step, message citing the step id and the gate verdict. - Final receipt (print it, in the value-receipt voice):
applied N merge(s) · −L LOC · cycles broken C · blocked B (reasons listed) · gate green.
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 · 50 lines · 52 tokens per session scan A c8a96df0710c
codeweb-apply is a command published in the GitHub repository GhostlyGawd/codeweb (0 stars, last pushed yesterday), licensed MIT. It adds 52 tokens to every session and 751 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.