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/yao-beyond/debug-hunter/debug-huntgit clone --depth 1 https://github.com/yao-beyond/debug-hunterWhat 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 | $0.00018 | $0.00530 |
| Opus 5 | $0.00009 | $0.00265 |
| Sonnet 5 | $0.00004 | $0.00106 |
| Haiku 4.5 | $0.00002 | $0.00053 |
Grade A, and why
debug-hunt 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
/debug-hunt
Run the debug-hunter closed loop on: $ARGUMENTS
You are the Tier-1 financial Debug Hunter orchestrator. Wearing both the correctness hat
(does the system miscalculate?) and the attacker hat (can an attacker make it calculate for
them?), execute the 7-stage loop defined in AGENTS.md:
THREAT-MODEL → DETECT → TRIAGE → REPRODUCE → FIX → VERIFY → GUARD + RECYCLE
How to run
- Load context — read
knowledge-base/KB-INDEX.mdfirst, then the meta/ground-truth/pattern files it points to, plusskills/debug-hunter/SKILL.md. When installed as a plugin, these live under${CLAUDE_PLUGIN_ROOT}/. - Stage 0 THREAT-MODEL — delegate to the
threat-modeleragent to map money flows and produce attack hypotheses for$ARGUMENTS. - Stage 1 DETECT (dual track) — run
detector(correctness) andsecurity-fraud-detector(taint source→sink) in parallel over$ARGUMENTS. - Stage 2 TRIAGE — score each finding (loss × probability × detection difficulty), assign P0–P3, apply forced-P0 rules.
- Stage 2.5 REPRODUCE — delegate to
reproducer; security findings need an attack PoC that violates a financial invariant. - Stage 3 FIX — delegate to
root-cause(5-Why + fix strategy + tests). - Stage 4 VERIFY — delegate to
verifier: "attack no longer succeeds ∧ invariants always hold". - Stage 5 GUARD + RECYCLE — delegate to
knowledge-writerto distil rules/invariants and re-run detection to confirm the class is now caught.
Only confirmed, evidence-backed findings may be raised as high severity. A finding stays "suspected" until its taint path, DB evidence, and counter-evidence check are complete.
If $ARGUMENTS is empty, ask the user which path or module to hunt in.
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 · 40 lines · 18 tokens per session scan A 587bb9648fa3
debug-hunt is a command published in the GitHub repository yao-beyond/debug-hunter (10 stars, last pushed 21d ago), licensed MIT. It adds 18 tokens to every session and 530 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.
Other commands, from other repositories
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.
specify
Create or update the feature specification from a natural language feature description.
analyze
Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.
converge
Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.
implement
Execute the implementation plan by processing and executing all tasks defined in tasks.md.