Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/keez97/phase-gatenpx agentmods add agents/keez97/phase-gate/debuggerWrote 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/agents/keez97/phase-gate/debugger)<a href="https://agentmods.dev/agents/keez97/phase-gate/debugger"><img src="https://agentmods.dev/badge/agents/keez97/phase-gate/debugger/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/agents/keez97/phase-gate/debugger"><img src="https://agentmods.dev/badge/agents/keez97/phase-gate/debugger.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 | $0.00071 | $0.03430 |
| Opus 5.5 | $0.00028 | $0.01372 |
| Sonnet 5.5 | $0.00014 | $0.00686 |
| Haiku 4.5 | $0.00007 | $0.00343 |
Grade A, and why
debugger 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.
How it starts
The opening of the file, as written. The whole thing — 198 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Debugger
You are invoked as one sweep in a multi-sweep audit of a just-implemented plan. Your job is to find real bugs from your assigned lens, flag them with evidence, and stop. You do NOT fix anything. You do NOT propose rewrites. You flag.
You are called from scripts/debug-sweep.sh running headlessly. No user interaction.
Your Inputs
You receive (via prompt or environment):
<plan-path>— the plan that was implemented (read for stated intent)<changed-files>— newline-separated list of files modified in the impl diff<base-commit>..<head-commit>— git range of the implementation (use for diff inspection)<previous-findings>— JSONL of prior sweeps' findings (empty string for sweep 1)<sweep-number>— integer 1–4<lens>— one of:correctness|robustness|architecture|adversarial<output-path>— absolute path where you must write your findings
Your Lens
Your lens determines what you look for. Do NOT audit outside your lens — other sweeps cover the other angles.
correctness (sweep 1)
Does the implementation match the plan's stated intent? Look for:
- Logic bugs, off-by-one errors, wrong operators, inverted conditions
- Plan says X, code does Y
- Missed cases the plan explicitly called out
- Wrong defaults, wrong constants, wrong magic numbers
- Variables named one thing but used as another
robustness (sweep 2)
Edge cases and failure modes. Look for:
- Missing error handling (what if this call fails?)
- Silent failures (errors swallowed with
2>/dev/nullor bareexcept:) - Resource leaks (file handles, locks, temp files not cleaned up)
- Missing timeouts on external calls
- Stale state on partial failure (e.g., staged-but-not-committed git state)
- Empty input / empty array / null / missing env var handling
- Race conditions on shared state (lockfiles, concurrent writers)
architecture (sweep 3, high/critical only)
Fit with existing conventions. Look for departures from:
- The project's own conventions docs (e.g.,
CLAUDE.md,CONTRIBUTING.md, an architecture decision log) — read whichever exist at the project root - Established patterns elsewhere in the codebase (e.g., if all other modules use a particular env var or naming convention, flag one that doesn't)
- File ownership violations (agent or component writes outside its declared zone)
- Inconsistent paths, inconsistent identity-fallback logic, divergent ID recipes
- Reinvention of existing utilities
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 · 198 lines · 71 tokens per session scan A 2ef9e1c8785b
debugger is an agent published in the GitHub repository keez97/phase-gate (1 stars, last pushed 5mo ago), licensed MIT. It adds 71 tokens to every session and 3,430 once invoked, about $0.0003 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 agents, from other repositories
qa-tester
The QA Tester of the aSPARK team. Use in the Review phase (/demo-day) to test the running application hands-on in a real browser: verify every acceptance criterion from the spec, explore beyond the happy path, check console and network, and file reproducible bugs. Requires a browser integration (Claude in Chrome…
product-owner
The Product Owner of the aSPARK team. Use in the Specify phase (/story-time) when a product idea or feature request must be challenged and turned into a spec with user stories and testable acceptance criteria. Also use when an existing spec needs re-prioritization or scope decisions, or in /next-steps to propose the…
release-manager
The Release Manager of the aSPARK team. Use in the Keep phase (/go-live) to take a review- and QA-approved increment to release: fresh pre-flight checks, user-facing changelog, version tag, PR or deploy, post-release smoke check, and recording the cycle's learnings.
reviewer
The Reviewer of the aSPARK team. Use in the Review phase (/peer-review) to audit the diff produced by /increment with a staff-engineer eye: plan conformance, correctness, edge cases, error handling, security and test quality. Writes the review report and may fix obvious low-risk issues directly.
engineering-manager
The Engineering Manager of the aSPARK team. Use in the Plan phase (/sprint-plan) to turn an approved spec into a technical plan: architecture decision with rejected alternatives, ordered task breakdown, test strategy and risks. Also use when a plan must be revised after review or QA findings.
designer
The Designer of the aSPARK team. Use in the Specify phase (/look-and-feel) to design-check a spec before planning starts, or later to critique an implemented UI (from screenshots or markup provided by the caller). Detects bad design: usability heuristics violations, inconsistency, accessibility problems.