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/cognitx-leyton/codegraph/trace-endpointgit clone --depth 1 https://github.com/cognitx-leyton/codegraphWhat 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 | $0.00014 | $0.00691 |
| Opus 5 | $0.00007 | $0.00345 |
| Sonnet 5 | $0.00003 | $0.00138 |
| Haiku 4.5 | $0.00001 | $0.00069 |
Grade A, and why
trace-endpoint 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 2d ago.
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 — 60 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Usage
/trace-endpoint <url_substring>
Matches against :Endpoint.path CONTAINS $ARGUMENTS — so partial matches work (/users picks up /users/:id, /users/profile, etc.).
Runs two queries in sequence:
- Surface — the matched endpoints and their handler methods (the API shape).
- Reach — every method reachable within 4
CALLShops from each handler, grouped by enclosing class (the data-and-compute reach).
Useful for impact analysis ("what breaks if I change this URL?"), security review ("what does this public endpoint touch?"), and onboarding ("show me a feature end-to-end").
What this does
codegraph query "
MATCH (m:Method)-[:HANDLES]->(e:Endpoint)
WHERE e.path CONTAINS '$ARGUMENTS'
OPTIONAL MATCH (ctrl:Class)-[:HAS_METHOD]->(m)
RETURN 'surface' AS layer,
e.method AS http_method,
e.path AS route,
ctrl.name AS controller,
m.name AS handler,
e.file AS file
ORDER BY e.path
LIMIT 20
"
codegraph query "
MATCH (m:Method)-[:HANDLES]->(e:Endpoint)
WHERE e.path CONTAINS '$ARGUMENTS'
MATCH path = (m)-[:CALLS*1..4]->(target:Method)
MATCH (enclosing:Class)-[:HAS_METHOD]->(target)
WITH DISTINCT enclosing.name AS class, target.name AS method, target.file AS file
RETURN 'reach' AS layer, class, method, file
ORDER BY class, method
LIMIT 100
"
Caveats
- Python has no
:Endpointnodes yet. Stage 1 Python only captures classes / functions / imports / decorators; FastAPI / Flask / Django route detection lands in Stage 2. Running this against a URL from a Python-only project returns zero rows. Document the gap rather than pretending the graph has coverage it doesn't. - Only typed CALLS resolve reliably. Python's name-resolved calls (
obj.foo()whereobj's type isn't known) may miss. TS:this.x()resolves; dynamic dispatch may not. - 4-hop bound is a trade-off. Deep call chains (e.g. orchestrator → saga → worker → DB) may exceed it. Raise to
*1..6if you're doing deep security review — slower but more complete. CALLS_ENDPOINTis the inverse — that's the edge for outgoing HTTP calls to an endpoint (e.g. frontend fetching from backend). If you want "who hits this endpoint from elsewhere", queryCALLS_ENDPOINTinstead.
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.
- 2d ago First seen · 60 lines · 14 tokens per session scan A ab4c9375ef8c
trace-endpoint is a command published in the GitHub repository cognitx-leyton/codegraph (11 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 14 tokens to every session and 691 once invoked, about $0.0001 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 2026-08-30.
Other commands, from other repositories
feedback
Security Design Review — PRD/기획서 기반 보안 의견서·검토 의견서 생성.
compliance
이전 보안 진단 보고서의 Finding들이 패치되었는지 확인하고, 변경된 코드에서 신규 취약점을 탐색합니다.
va
Vulnerability Assessment — 8차원 아키텍처 진단 + Self-Verify + Evidence Verification.
pentest
Penetration Testing — 시나리오 기반 모의해킹 + POC + 라이브 검증 (State Delta 기반).
redteam
Red Team Operations — 인프라 설정 보안 리뷰 + MITRE ATT&CK + Detection Engineering.
run
CH015 CISO에게 보안 진단을 지시합니다. 자연어로 요청하면 적절한 Division과 에이전트를 자동 할당합니다.