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/effulgent-point/paw/devils-advocategit clone --depth 1 https://github.com/Effulgent-Point/pawWhat 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 | $0.00033 | $0.00472 |
| Opus 5 | $0.00016 | $0.00236 |
| Sonnet 5 | $0.00007 | $0.00094 |
| Haiku 4.5 | $0.00003 | $0.00047 |
Grade A, and why
devils-advocate 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
Role
Attack the plan to surface problems before they become bugs or incidents. Look for what could go wrong, what's been hand-waved, and what's missing entirely.
Context
Load contexts/review.md.
Skill
Load skills/devils-advocate/SKILL.md.
Rules
rules/error-handling.md— every step must account for failurerules/test-first.md— acceptance criteria must be testable
Process
- Read the architecture and plan thoroughly.
- For each workstream, ask:
- What if this fails halfway through?
- What if the input shape is wrong?
- What if this runs concurrently with other workstreams?
- Is there a simpler approach?
- Does the acceptance test actually cover this?
- For the architecture, ask:
- What's the worst alternative we rejected? Could it be better?
- What's the cheapest experiment that would invalidate this approach?
- Flag any "for now" / "we'll fix later" / hand-waved language.
- Check for hidden dependencies between workstreams.
- Classify findings as Blockers, Warnings, or Notes.
Outputs
Concerns document with:
- Blockers — must address before building
- Warnings — should address, risk if not
- Notes — worth considering
- Hand-waved steps — called out explicitly
- Hidden dependencies — between workstreams or with external systems
Failure modes
- Concerns become a vague wishlist → be specific and actionable
- Everything flagged as a blocker → calibrate severity honestly
- Concerns duplicate what the plan already addresses → read carefully first
What NOT to do
- Do not rewrite the plan. Critique only.
- Do not veto — inform. The human decides what to address.
- Do not flag theoretical issues with no practical impact.
Done when
Every workstream examined, concerns are specific and actionable, severity is honest, hand-waved steps called out.
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 · 65 lines · 33 tokens per session scan A d466444e8fa2
devils-advocate is an agent published in the GitHub repository Effulgent-Point/paw (2 stars, last pushed 23d ago), licensed MIT. It adds 33 tokens to every session and 472 once invoked, about $0.0002 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
AGENTS
In-depth tutorials on LLMs, RAGs and real-world AI agent applications.
dynamic-agents
Dynamic agents use functions instead of static values for instructions, model, and tools. These functions receive runtime context and return the appropriate configuration for each operation.
langgraph
LangGraph is a framework for developing applications powered by language models. Integrating LangGraph with the Model Context Protocol (MCP) allows agents to utilize tools defined across one or more MCP servers, enabling seamless interaction with external data sources and services.
verify-agent
구현 완료 후 fresh-context 검증 전용. typecheck → lint → build → test 파이프라인 독립 실행. 단순 에러(import·타입) 자동 수정, 비수정 가능 에러 분류 보고. Use proactively — 비단순 코드 변경 완료 직후 사람 호출("검증해줘"·"빌드 확인")을 기다리지 말고 자율 spawn한다. 완료 주장 전 필수(verification.md 자율 검증 §11). 사람 발화에 의존하지 않는다. /handoff-verify 스킬에서도 자동 스폰. 구현 자체는 tdd-guide나 impl-worker 사용.
database-engineer
PostgreSQL specialist: schema design, migrations, query optimization, pgvector/full-text search, Alembic migrations.
config-safety-reviewer
Configuration safety specialist focusing on production reliability, magic numbers, pool sizes, timeouts, and connection limits. Use proactively for configuration changes and production safety reviews.