The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were identified. The skill follows best practices by proposing read-only analysis and offering recommendations for user-implemented safeguards (such as blocking edits to .env files).
[COMMAND_EXECUTION]: The skill uses Bash for benign codebase inspection. It executes commands like ls -la and cat package.json to identify project dependencies, structure, and existing configurations. These operations are limited to metadata and non-sensitive configuration files.
[PROMPT_INJECTION]: The skill has an inherent surface for indirect prompt injection as it processes untrusted codebase data (e.g., project configuration files and directory structures) to generate its reports.
Ingestion points: Phase 1 (Codebase Analysis) in SKILL.md reads package.json, pyproject.toml, and other project files.
Boundary markers: No explicit delimiters are used to isolate codebase content from the analysis logic.
Capability inventory: The skill utilizes Read, Glob, Grep, and Bash for its analysis functions.
Sanitization: No explicit sanitization or filtering of the read file content is mentioned before it is processed for recommendations.

claude-automation-recommender