Claude Code Spec Workflow is a set of Claude Code commands and agents that organizes software work into documented stages for new features and bug fixes. It is for developers who want Claude Code to handle requirements, design, implementation, issue resolution, and verification through repeatable workflows, with the catalogue entries providing the associated commands and agents.
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.
git clone --depth 1 https://github.com/elementalsouls/Claude-BugHunterWrote 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/commands/elementalsouls/claude-bughunter/autopilot)<a href="https://agentmods.dev/commands/elementalsouls/claude-bughunter/autopilot"><img src="https://agentmods.dev/badge/commands/elementalsouls/claude-bughunter/autopilot/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/commands/elementalsouls/claude-bughunter/autopilot"><img src="https://agentmods.dev/badge/commands/elementalsouls/claude-bughunter/autopilot.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 | $0.00049 | $0.00933 |
| Opus 5.5 | $0.00020 | $0.00373 |
| Sonnet 5.5 | $0.00010 | $0.00187 |
| Haiku 4.5 | $0.00005 | $0.00093 |
Grade A, and why
autopilot 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.
Copies of this mod
1 near-identical copy found in the catalogue:
- autopilot — 100% identical, 0 lines differ
How it starts
The opening of the file, as written. The whole thing — 96 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/autopilot
Autonomous hunt loop with deterministic scope safety and configurable checkpoints.
Usage
/autopilot target.com # default: --paranoid mode
/autopilot target.com --normal # batch checkpoint after validation
/autopilot target.com --yolo # minimal checkpoints (still requires report approval)
/autopilot target.com --quick # fast surface scan, fewer checks, lower token use
/autopilot targets.txt # multiple targets — one domain per line in the file
Session Isolation (Important)
Start a fresh Claude Code session per target. Claude accumulates context across a session — testing multiple targets in one session causes cross-contamination where findings, payloads, and tech stack assumptions from target A bleed into target B.
Best practice:
# Terminal 1: target A
claude → /autopilot targetA.com
# Terminal 2: target B (separate process)
claude → /autopilot targetB.com
If you must test multiple targets in one session, run /pickup target.com at the start of
each target switch to reload the correct context.
Token Optimization
Use --quick for faster, lower-cost scans (skips deep fuzzing and extended nuclei templates):
/autopilot target.com --quick # ~40% fewer tokens, covers main attack surface
/hunt target.com --vuln-class idor # single bug class — lowest token use
For long hunts, run /compact (Claude Code built-in) periodically to compress context
without losing findings.
What This Does
Runs the full hunt cycle without stopping for approval at each step:
1. SCOPE Load and confirm program scope
2. RECON Run recon (or use cached if < 7 days old)
3. RANK Prioritize attack surface (recon-ranker agent)
4. HUNT Test P1 endpoints systematically
5. VALIDATE 7-Question Gate on findings
6. REPORT Draft reports for validated findings
7. CHECKPOINT Present to human for review
Safety Guarantees
- Every URL is checked against the scope allowlist before any request
- Reports are NEVER auto-submitted — always requires explicit approval
- PUT/DELETE/PATCH require human approval in --yolo mode (safe methods only)
- Circuit breaker stops hammering if 5 consecutive 403/429/timeout on same host
- Rate limited at 1 req/sec (testing) and 10 req/sec (recon)
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 · 96 lines · 0 tokens per session scan A 2fe48b49beb1
autopilot is a command published in the GitHub repository elementalsouls/Claude-BugHunter (4,874 stars, last pushed 6d ago), licensed MIT. It adds 49 tokens to every session and 933 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 2026-08-30.
Other commands, from other repositories
web3-audit
Smart contract security audit — runs through 10 bug class checklist (accounting desync, access control, incomplete path, off-by-one, oracle errors, ERC4626, reentrancy, flash loan, signature replay, proxy/upgrade). Applies pre-dive kill signals first. Generates Foundry PoC template for confirmed findings. Usage…
bypass-403
Probe a 403/401 endpoint with the most-paid bypass tricks (header injection, path encoding, method swap, WAF fingerprint, vendor-specific). Wraps byp4xx when installed; otherwise runs a built-in matrix of 38+ techniques. Usage: /bypass-403 | /bypass-403 -l.
spray
Password spray with hard guards — typed-hostname confirmation, lockout warning, audit log. Modes: http-form (custom login page), oauth (password grant), o365 + okta (via TREVORspray). Default delay 30min/round + 60s jitter. Usage /spray --mode --users --passes.
autopilot
Run autonomous hunt loop on a target — scope check → recon → rank surface → hunt → validate → report with configurable checkpoints. Usage: /autopilot target.com [--paranoid|--normal|--yolo].
breach-check
HIBP k-anonymity check on a password wordlist. Enriches each password with its breach count, ranks DESC. Free API (no key), only first 5 chars of SHA-1 sent. Output -> -ranked.txt. Usage /breach-check [--min-count N] [--max-count N] [--with-counts].
recon
Run the full recon pipeline by invoking tools/reconengine.sh — subdomain enum (subfinder + amass + crt.sh + wayback), httpx live host probing with tech detection, nmap port scan, gau URL collection, JS analysis, ffuf directory fuzzing, parameter discovery, config exposure check, CI/CD workflow scan. Outputs to recon/…