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 atretyak1985/swarmery --skill auditing-agent-architecturegit clone --depth 1 https://github.com/atretyak1985/swarmeryWrote 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/atretyak1985/swarmery/auditing-agent-architecture)<a href="https://agentmods.dev/skills/atretyak1985/swarmery/auditing-agent-architecture"><img src="https://agentmods.dev/badge/skills/atretyak1985/swarmery/auditing-agent-architecture/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/atretyak1985/swarmery/auditing-agent-architecture"><img src="https://agentmods.dev/badge/skills/atretyak1985/swarmery/auditing-agent-architecture.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 | $0.00118 | $0.01327 |
| Opus 5 | $0.00059 | $0.00664 |
| Sonnet 5 | $0.00024 | $0.00265 |
| Haiku 4.5 | $0.00012 | $0.00133 |
Grade A, and why
auditing-agent-architecture 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 6d 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 — 101 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Auditing Agent Architecture
Overview
Best-practice reference for the architecture layer of Claude-based agents: the execution loop, orchestration topology, subagent contracts, enforcement gates, hooks, decomposition, and session state. Each reference module states the correct pattern, the named anti-patterns, and ends with an Audit Checklist of verifiable conditions.
When to use
- Reviewing or designing an agent loop, coordinator, or subagent structure
- Debugging: premature termination, infinite loops, dropped tool results, stalled or repeating agents, coverage gaps between subagents
- Deciding: prompt guidance vs hard gates, hooks vs prompts, fixed pipeline vs dynamic decomposition
Not for tool schemas or MCP servers (use designing-tools-and-mcp) or context-window degradation (use managing-context-reliability).
Quick reference
| Module | Read when the question is about |
|---|---|
references/1-1-agentic-loops.md |
Loop lifecycle, stop_reason branching, termination anti-patterns |
references/1-2-multi-agent-orchestration.md |
Hub-and-spoke coordination, context isolation, decomposition coverage gaps |
references/1-3-subagent-invocation-and-context-passing.md |
Task tool, structured metadata handoff, parallel spawning, fork_session |
references/1-4-workflow-enforcement-and-handoff.md |
Prompt guidance vs programmatic gates, prerequisite gates, handoff protocols |
references/1-5-agent-sdk-hooks.md |
PreToolUse/PostToolUse policy enforcement, hooks vs prompt instructions |
references/1-6-task-decomposition-strategies.md |
Fixed sequential vs dynamic decomposition, attention dilution |
references/1-7-session-state-and-resumption.md |
Session persistence, stale context, targeted re-analysis |
How to audit
- Match the system under review to modules in the table; read those files.
- Run each module's Audit Checklist item by item against the actual code/config.
- Report every unchecked item as a finding with
file:lineevidence and the module's recommended fix.
What ships with it
8 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- checklists.md 9.0 KB
- references/1-1-agentic-loops.md 7.2 KB
- references/1-2-multi-agent-orchestration.md 8.1 KB
- references/1-3-subagent-invocation-and-context-passing.md 8.4 KB
- references/1-4-workflow-enforcement-and-handoff.md 10 KB
- references/1-5-agent-sdk-hooks.md 8.3 KB
- references/1-6-task-decomposition-strategies.md 9.0 KB
- references/1-7-session-state-and-resumption.md 9.5 KB
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.
- 6d ago Changed · +32 tokens per session 3c1e32408f00
- 11d ago First seen · 101 lines · 86 tokens per session scan A b0290afd5a66
auditing-agent-architecture is a skill published in the GitHub repository atretyak1985/swarmery (5 stars, last pushed yesterday), licensed Apache-2.0. It adds 118 tokens to every session and 1,327 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 skills, from other repositories
notebooklm
Programmatic access to Google NotebookLM for creating notebooks, adding sources (URLs, YouTube, PDFs, audio, video, images), chatting with content, generating artifacts (podcasts, videos, reports, quizzes, mind maps, flashcards, infographics), and downloading results. Includes capabilities beyond the web UI.
gh-address-comments
This skill should be used when addressing GitHub pull request review comments systematically with mandatory resolution tracking. Use when user says "address PR comments", "fix review feedback", "handle PR review", "resolve PR threads", or needs to ensure all review comments are addressed before merging. Fetches…
prowlarr
Search indexers and manage Prowlarr. Use when the user asks to "search for a torrent", "search indexers", "find a release", "check indexer status", "list indexers", "prowlarr search", "sync indexers", or mentions Prowlarr/indexer management.
sonarr
This skill should be used when managing TV shows in Sonarr. Use when the user asks to "add a TV show", "search Sonarr", "find a series", "add to Sonarr", "remove a show", "check if show exists", "Sonarr library", "TVDB lookup", or mentions TV show management or Sonarr operations.
commit-message
Use ANY time the user signals they want to commit staged changes — including bare "commit", "/commit", "commit this", "commit these changes", "let's commit", "make a commit", "git commit", "write a commit message", "make a conventional commit", "use conventional commits", or any equivalent phrasing in context (e.g.…
dependency-versions
MUST consult this skill before answering whenever the user's task involves external versioned dependencies — even if you think you can handle it directly. This applies to: checking if packages/tools are up to date, upgrading npm/pip/cargo/go dependencies, planning or writing CI/CD workflows (GitHub Actions, CircleCI…