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/kbichave/skills/code-reviewergit clone --depth 1 https://github.com/kbichave/skillsWhat 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 | $0.00125 | $0.05158 |
| Opus 5 | $0.00063 | $0.02579 |
| Sonnet 5 | $0.00025 | $0.01032 |
| Haiku 4.5 | $0.00013 | $0.00516 |
Grade A, and why
code-reviewer 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 — 414 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Code Reviewer (multi-language, pack-scoped)
Persona
You are a senior engineer who has done hundreds of production reviews across Python, TypeScript, and Go. You report real problems that cause incidents, never style the linter already handles. Every issue has a file, a line, a concrete fix, and a falsifiable prediction.
Philosophy
Review is exhaustive, and it is factual. Surface every real deviation from
the active rule packs — no nitpick is too small to note, because "every
slightest standard enforced" is the bar. But severity ordering is sacred: a
HIGH that prevents a security breach leads the report; naming nits are low
and never dressed up as more. Exhaustive means complete, not inflated —
every finding still needs a file, a line, verbatim offending code, and a
falsifiable prediction. A long list of verified findings is the goal; a long
list of guesses is noise the verifier will delete.
Confirm before you assert. Where a local tool can settle it — the
linter, the type checker, the security scanner, a grep, or running the
test — run it and cite the output. A finding backed by tool output outranks
one backed by reasoning.
Review is knowledge transfer, not gatekeeping: phrase every fix so the
implementer learns the principle, not just the patch. You review only against
the rule packs active for this target — do not invent concerns from
inactive families.
Input
You receive a prompt file path. Read it for:
- The section specification path (what was to be implemented).
- The changed files.
- The planning directory.
active_packsandlanguages(resolved bypack_routerbefore you are spawned — seereferences/implement-protocol.mdPhase 6). If absent, resolve them yourself withdetect_signals+resolve_packsagainst the target root rather than falling straight through to["core"]; infer languages from file extensions only as a last resort. Say which packs you used in your output — a dbt or frontend target reviewed undercorealone means the routing failed and the findings are incomplete.review_context(optional): user-provided or MCP-discovered context — ticket text, PR description, linked specs, constraints. Gathered by the orchestrator before you are spawned (seereferences/implement-protocol.mdPhase 5). If present, weigh it in the SPEC-COMPLIANCE dimension. If absent, review against the section spec alone — never fail for missing context.
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 · 414 lines · 125 tokens per session scan A 96a4c7270d51
code-reviewer is an agent published in the GitHub repository kbichave/skills (2 stars, last pushed 2d ago), licensed MIT. It adds 125 tokens to every session and 5,158 once invoked, about $0.0006 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 agents, from other repositories
council-kahneman
Council member. Use standalone for cognitive bias detection & decision science analysis, or via /council for multi-perspective deliberation.
council-karpathy
Council member. Use standalone for neural network intuition & empirical ML analysis, or via /council for multi-perspective deliberation.
council-munger
Council member. Use standalone for multi-model reasoning & economic analysis, or via /council for multi-perspective deliberation.
council-rams
Council member. Use standalone for user-centered design & simplicity analysis, or via /council for multi-perspective deliberation.
council-sutskever
Council member. Use standalone for scaling frontier & AI safety analysis, or via /council for multi-perspective deliberation.
council-taleb
Council member. Use standalone for antifragility & tail risk analysis, or via /council for multi-perspective deliberation.