The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill processes data from the repository and git history to calculate maturity scores. While this presents an attack surface where malicious content in files could attempt to influence the agent, the skill implements comprehensive mitigations:
Ingestion points: Data is ingested in SKILL.md via the repo-scout agent and git log exports.
Boundary markers: While explicit delimiters are not used, the instructions clearly define the data as untrusted and separate it from execution logic.
Capability inventory: The skill can perform file system operations (git log, mktemp, rmdir) and write HTML reports as described in SKILL.md.
Sanitization: All data from repo-scout is validated against strict regex patterns (e.g., path and link verification), and any data failing these checks causes the entire set to be discarded. All output in the HTML report is HTML-escaped.
[Safe Command Execution]: The skill uses shell commands for standard operational tasks like managing temporary directories and querying git history. These operations are gated by user consent and do not interpolate untrusted user input directly into shell execution strings, minimizing the risk of command injection.
[Data Privacy and Protection]: The skill includes explicit instructions to avoid sensitive data, such as secrets or individual-level metrics. It focuses on organizational-level metadata and enforces a restrictive Content Security Policy (CSP) in the generated HTML to prevent external data exfiltration.

maturity