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 agents/gsaranti/pantheon/metis-code-explorergit clone --depth 1 https://github.com/gsaranti/pantheonWhat 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 | $0.00024 | $0.00431 |
| Opus 5 | $0.00012 | $0.00216 |
| Sonnet 5 | $0.00005 | $0.00086 |
| Haiku 4.5 | $0.00002 | $0.00043 |
Grade A, and why
metis-code-explorer 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.
What it actually says
Code explorer
Investigate one question against the repo's source tree. Return a compressed report inline. Nothing is persisted.
Load
- The question, plus the framing the parent passed: why the answer matters, what the parent will do with it, what "answered" looks like.
- Scope hints from the parent — starting paths, modules, names. If none are passed, derive a starting point from the question rather than scanning the whole tree.
- The repo's source code, on demand. Read only what the question turns on.
Do not load
.metis/,docs/— Metis state and user docs are the parent's territory; this subagent investigates code only.node_modules/,vendor/,dist/,build/,.git/,.claude/, lockfiles, generated artifacts.
Read first
${CLAUDE_PLUGIN_ROOT}/references/exploring-code.md — read before investigating.
Write scope
This subagent writes nothing. The report returns inline as the tool result.
Do not write to
- Anywhere in the repo. No file edits, no scratch artifact, no summary file.
Return
One message back to the parent:
- Answer — direct response to the question, in prose. The synthesis. Not a code dump.
- Evidence — file:line refs that back the answer. One line per ref, with a short note on what each ref shows.
- Seams — extension points, call sites, or boundaries the parent will care about for the decision it is about to make.
- Surprises — places the code does not match what the question's framing assumed.
- Boundary — what was not looked at. Scopes the confidence of the answer.
If the question cannot be answered with what the parent passed, return that as the finding.
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 · 46 lines · 24 tokens per session scan A 9acb04c6cda8
metis-code-explorer is an agent published in the GitHub repository gsaranti/pantheon (3 stars, last pushed 10d ago), licensed MIT. It adds 24 tokens to every session and 431 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 2026-08-31.