Borrowing it
Nothing to install: this file belongs to MartyBonacci/specswarm. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/MartyBonacci/specswarm/main/CLAUDE.mdgit clone --depth 1 https://github.com/MartyBonacci/specswarmWrote 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/instructions/martybonacci/specswarm/claude-md)<a href="https://agentmods.dev/instructions/martybonacci/specswarm/claude-md"><img src="https://agentmods.dev/badge/instructions/martybonacci/specswarm/claude-md/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/instructions/martybonacci/specswarm/claude-md"><img src="https://agentmods.dev/badge/instructions/martybonacci/specswarm/claude-md.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 | $0.02691 | $0.02691 |
| Opus 5.5 | $0.01076 | $0.01076 |
| Sonnet 5.5 | $0.00538 | $0.00538 |
| Haiku 4.5 | $0.00269 | $0.00269 |
Grade A, and why
specswarm CLAUDE.md 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 2mo 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 — 113 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md — SpecSwarm
Overview
SpecSwarm is a Claude Code plugin providing spec-driven development workflows: Build, Modify, Fix, Ship. As of v6.0.0, all functionality lives in the ss plugin and is invoked via /ss:* commands (23 visible + 11 internal/hidden = 34 total — v7.1.0 /ss:preflight, v7.2.0 /ss:notify, v7.3.0 /ss:intervention, v7.4.0 /ss:verify, v7.5.0 /ss:retrospective, v7.6.0 /ss:decisions, v7.8.0 /ss:dry-run, v7.9.0 /ss:watchdog, v7.10.0 /ss:overnight, v7.18.0 /ss:conduct + /ss:mentor-init). v7.11.0 hardens the verification loop (WARN-on-zero preflight gates, canonical-checkbox /ss:tasks, verify-queue drain in /ss:implement, ship verify-queue precondition, decision-miner prior-commitments input). v7.12.0 collapses the verify-queue Stop hook into a per-feature summary when pending > SPECSWARM_VERIFY_QUEUE_LIST_MAX (default 8), so large queues stop flooding the conversation. v7.13.0 (AUTO-MAGIC epic Phase 2 — see docs/designs/automagic-epic-plan.md) adds the taste model (lib/taste.sh, memory corpus as distilled-rulings store with check-type: frontmatter, accreting from decisions/clarify/verify answers) and assume-first clarify (three-way gap ladder, structured provenance-cited ## Assumptions ledger, batched fork questions, assumptions-provenance preflight check). v7.14.0 adds blocking per-slice build+lint gates (SPECSWARM_SLICE_GATES, real test-framework-detector.sh/quality-gates.sh), verify-time diff screeners (geometry/roundtrip/fixture-shape/test-globs → spec-mentor clauses), and the NEEDS-SIGHTED queue state that blocks /ss:ship until human sighted sign-off (/ss:verify --sighted or ship-time batch). v7.15.0 closes the distiller loop (deterministic rulings auto-generate edit-time hooks via embedded specswarm-rule blocks; spec-mentor receives judgment taste rules in its bundle) and hardens headless runs (sync/budget clauses in every dispatch, work-tree-aware failure classification, one-shot auto-resume — lib/overnight/resilience.sh). v7.16.0 adds /ss:bakeoff (calibration loop: N candidates → contact sheet → verdict → pinned winner + distilled ruling), [DECIDED-BY-DATA: metric, review-when] fork-deferral markers (tracked by metrics + watchdog), and the ground-truth-bias preflight check (7th). v7.17.0 caps the epic with /ss:go — the single default entry point walking the full ladder via hooks/go-loop-hook.sh + .specswarm/go-loop.state, pausing only at 4 human touchpoints (assumptions review, decision sheet, sighted gates, ship blessing); /ss:build untouched. v7.18.0 extracts the proven mentor→builder loop (four-minds pilot: ~500 audited dispatches) into /ss:conduct (hardened headless dispatch via lib/conduct/dispatch.sh — per-tree flock, process-group timeout kill, survivor + orphan-listener sweeps, opus-pinned, audited runs at .specswarm/conduct/runs/ — plus the independent-verification protocol and six failure-mode resume recipes), /ss:mentor-init (mentor-directory scaffold from templates/mentor/ with a SessionStart role hook), and watchdog conduct-run monitoring (auto-enabled when .specswarm/conduct/ exists). Includes 10 natural-language skills, 6 multi-agent orchestration agents with v7.7.0 explicit model assignments (5 on opus, 1 on haiku), and the .specswarm/ per-project state directory (directory name preserves the SpecSwarm brand).
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.
- 2mo ago First seen · 113 lines · 2,691 tokens per session scan A a58663b00bb7
specswarm CLAUDE.md is an instructions file published in the GitHub repository MartyBonacci/specswarm (67 stars, last pushed 3mo ago), licensed MIT. It adds 2,691 tokens to every session, about $0.0108 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 2026-08-30.
specswarm CLAUDE.md compared with instructions from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| specswarm CLAUDE.mdMartyBonacci/specswarm | Instructions file | 67 | 3mo ago | MIT | A | 2,691 | Claude Code |
| vscode buildNext.instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 9,719 | GitHub Copilot |
| next.js AGENTS.mdvercel/next.js | Instructions file | 143k | yesterday | MIT | A | 7,461 | Codex, OpenCode |
| langchain AGENTS.mdlangchain-ai/langchain | Instructions file | 149k | yesterday | MIT | A | 4,680 | Codex, OpenCode |
| vscode oss-third-party-notices.instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 5,516 | GitHub Copilot |
| vscode copilot-instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 3,093 | GitHub Copilot |
| deepseek-harness AGENTS.mddeepseek-ai/deepseek-harness | Instructions file | 274k | 2d ago | MIT | A | 4,313 | Codex, OpenCode |