Bug scan

Skill caminoezequiel/ai-tools/skills/bug-scan

Multi-stage bug scanning pipeline for Node.js codebases. Runs 7 specialized agents — Surface Analyzer, Flow Tracer (per entry point), Context Compiler, Bug Hunter, Bug Evaluator, Bug Judge, and Report Formatter — to map, trace, validate, and prioritize real bugs using data flow analysis. Use when asked to scan a path, module, feature, or full project for bugs, code smells, and security risks.From its SKILL.md

Install

npx -y skills add caminoezequiel/ai-tools --skill bug-scan

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its file declares

Copied from the file, not written here

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

3.2 KB, 612 tokens by cl100k_base, as published. Nobody here has run it

You have been invoked with: /bug-scan $ARGUMENTS

The target path is: $ARGUMENTS

If no path was provided, ask the user: "Which path should I scan?" and wait for their answer before proceeding.

Execute the following pipeline in strict order. Do not skip stages. Do not merge stages.

Phase 1 — Surface Analyzer

Use the surface-analyzer subagent. Pass it: the path $ARGUMENTS. Wait for its full Surface Analyzer Report before proceeding.

HALT check: If the output contains SURFACE_ANALYZER_HALT, stop immediately and display the halt message to the user. Do not proceed to Phase 2.

Phase 2 — Flow Tracing (Supervisor Loop)

Extract the numbered entry point list from the Surface Analyzer Report.

If the entry point list is empty, output:

No entry points detected. The scan cannot continue. Check the Surface Analyzer Notes section for details.

Then stop.

Otherwise, for EACH entry point in the list, spawn one flow-tracer subagent. Pass it:

The full entry point string exactly as listed (e.g. src/routes/payments.ts:15 — express — POST /charge)
The original path $ARGUMENTS

Spawn workers sequentially — wait for each JSON output before spawning the next. Accumulate ALL JSON outputs. Do not discard any.

Phase 3 — Context Compiler

Use the context-compiler subagent. Pass it: all JSON outputs from Phase 2 concatenated in a single block. Wait for the compiled Markdown document (the Data Flow Map) before proceeding.

Phase 4 — Bug Hunter

Use the bug-hunter subagent. Pass it:

The Data Flow Map (full Markdown document from Phase 3)
The original path $ARGUMENTS

Wait for the complete JSON output before proceeding.

Phase 5 — Bug Evaluator

Use the bug-evaluator subagent. Pass it: the Bug Hunter's JSON output + the original path $ARGUMENTS. Wait for its JSON output (excluded + confirmed arrays) before proceeding.

Phase 6 — Bug Judge

Use the bug-judge subagent. Pass it: the Bug Evaluator's full JSON output + the original path $ARGUMENTS. Wait for the final JSON report before proceeding.

Phase 7 — Report Formatter

Use the report-formatter subagent. Pass it:

The original path $ARGUMENTS
The Bug Judge's full JSON output

Wait for the Markdown report.

Final output

Present the Markdown report produced by the Report Formatter exactly as returned — do not summarize or reformat it.

Also tell the user:

Which entry points had the most confirmed findings
Which trail breaks represent the highest unresolved risk
That the raw JSON report was saved to $ARGUMENTS/bug-scan-report.json

What ships with it: 7 files

23.4 KB alongside SKILL.md