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 skills add AnmarHani/SwarmVault --skill swarm-debuggit clone --depth 1 https://github.com/AnmarHani/SwarmVaultWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/anmarhani/swarmvault/swarm-debug)<a href="https://agentmods.dev/skills/anmarhani/swarmvault/swarm-debug"><img src="https://agentmods.dev/badge/skills/anmarhani/swarmvault/swarm-debug/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/anmarhani/swarmvault/swarm-debug"><img src="https://agentmods.dev/badge/skills/anmarhani/swarmvault/swarm-debug.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00052 | $0.00511 |
| Opus 5.5 | $0.00021 | $0.00204 |
| Sonnet 5.5 | $0.00010 | $0.00102 |
| Haiku 4.5 | $0.00005 | $0.00051 |
Grade A, and why
swarm-debug 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 1mo 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
swarm-debug — systematic debugging
No fix without a reproduced cause. Shotgun patching is how the same bug costs twice.
The loop
- Reproduce — a command/test that shows the bug, or record explicitly why reproduction is impossible. This is the entry gate for everything below.
- Evidence — read the FULL error, stack, and surrounding logs (not the first line); note what changed recently (git log, new deps, config).
- Locate — binary-search the causal chain: which layer last had correct data? Instrument with targeted prints/asserts if needed; remove them after.
- Hypothesize — ranked list, cheapest-to-test first. State each as a falsifiable claim ("the cache returns stale X when Y").
- Verify — one targeted experiment per hypothesis. Evidence kills or confirms; opinion does neither.
- Fix the root cause — not the symptom, when the root is reachable. If you must ship a symptom patch, say so in the code-note and open a ticket for the root.
- Regression test — must fail on the pre-fix code, pass on the fix. No test, no fix.
- Record (J1) — non-trivial diagnosis → memory note (the footgun, compact,
≤ 15 lines) + code-note link on the affected file, so no future session pays for this
twice. Check the vault FIRST next time:
swarmvault.py query --search "<symptom>"— someone may already have paid.
Hard rules
- Three dead hypotheses → stop, step back, re-examine assumptions — including the spec and design ("is the requirement itself contradictory?").
- A diagnosis that reveals a spec/design gap → route it to the question queue / SRS change management (swarm-spec), don't silently patch around it.
- Performance bugs: measure before and after; a fix without numbers is a guess.
Influences: superpowers systematic-debugging; Pocock's diagnosing-bugs; Jeffallan's debugging-wizard — see CREDITS.md.
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.
- 1mo ago First seen · 41 lines · 52 tokens per session scan A 739a2f4fdb4f
swarm-debug is a skill published in the GitHub repository AnmarHani/SwarmVault (5 stars, last pushed 2mo ago), licensed MIT. It adds 52 tokens to every session and 511 once invoked, about $0.0002 per session on Opus 5.5. 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 skills, from other repositories
agent-performance-optimizer
Agent skill for performance-optimizer - invoke with $agent-performance-optimizer.
agent-refinement
Agent skill for refinement - invoke with $agent-refinement.
langsmith-tracing
LangSmith tracing and debugging setup for LLM applications. Configure observability, capture traces, and enable debugging for LangChain/LangGraph agents.
behavior-contract
Bug condition/postcondition formalization as testable Behavior Contracts. Defines invariants that must be preserved across fixes.
quality-hooks
Language-specific auto-lint/format/typecheck pipeline. Supports Python (ruff+pyright), TypeScript (prettier+eslint+tsc), Go (gofmt+golangci-lint). Auto-fix and convergence loops.
aafb9b1b091b4dc00c7d838fa91f75c507e51c36045f31bccfc864a8532819f8
Use exact retained evidence.