code-polish

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted source code from a repository. This creates an attack surface where malicious instructions could be embedded in the code or comments to influence the agent's behavior.
  • Ingestion points: Source code files within a Git repository identified during the 'Fixed Scope' phase.
  • Boundary markers: The skill uses a 'resolved-scope' mechanism to limit the agent's context and prevent it from processing irrelevant or dangerous files.
  • Capability inventory: The agent is instructed to run formatters, linters, typechecks, and test suites (e.g., go test, cargo test, gradle, swift test).
  • Sanitization: The skill emphasizes high-confidence simplifications and requires evidence-based defect reporting, which helps maintain the integrity of the analysis process.
  • [COMMAND_EXECUTION]: To verify code changes, the skill instructs the agent to execute various build and test commands (e.g., go mod tidy, cargo clippy, ./gradlew, xcodebuild).
  • Scope: Command execution is restricted to existing project recipes and standard toolchains (Go, Rust, Swift, Kotlin/Android, Python, Node.js).
  • Constraint: The instructions explicitly warn against 'inventing' new tasks and emphasize using the repository's configured tools, which limits the risk of arbitrary command execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 07:09 AM