The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as its primary function is to ingest and analyze untrusted external codebases.
Ingestion points: The skill reads the entire target codebase using the Read, Grep, and Glob tools in Phase 1 and Phase 3.
Boundary markers: The instructions do not specify the use of strict delimiters or explicit 'ignore instructions' warnings when interpolating code snippets into reviewer prompts.
Capability inventory: The skill utilizes Bash, Agent, Write, and Edit tools. Specifically, Phase 5 (behavior-preservation verifier) instructs the agent to run the project's dev server or test harness, which involves executing code from the audited codebase.
Sanitization: No explicit sanitization or escaping of the ingested code content is mentioned before it is processed by the AI reviewers.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform structural evaluations (via tests/eval.sh) and to run codebase-specific tooling (such as knip, ts-prune, or project test suites) during the auditing and verification phases. This is intended behavior for a codebase auditor but grants the agent permission to execute commands on the local environment.

rigorous-review