The Agent Skills Directory

[SAFE]: The skill defines a structured engineering review process and does not exhibit any patterns of obfuscation, malicious command execution, or unauthorized data access.
[DATA_EXFILTRATION]: The skill reads local project source code and internal rule sets (e.g., TDD.md) to perform its evaluation. This access is consistent with its stated purpose and does not involve sending sensitive information to external servers.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool for benign tasks, such as generating timestamps for report file names via the date command.
[PROMPT_INJECTION]: The skill's function of reviewing external code and plans introduces a surface for indirect prompt injection. 1. Ingestion points: Project plans and source files read through Read, Grep, and Glob tools. 2. Boundary markers: No explicit delimiters are defined to isolate untrusted content from the system instructions. 3. Capability inventory: The skill has access to Bash, Write, and WebSearch tools. 4. Sanitization: There is no explicit sanitization or validation of the input data before it is processed by the agent.

sci-review