The Agent Skills Directory

[SAFE]: The skill provides instructional content and templates for documentation. No malicious patterns, bypass attempts, or obfuscation were identified across the nine reference files.\n- [NO_CODE]: The skill is composed entirely of Markdown reference files and configuration. It does not contain or execute scripts, binaries, or subprocesses, as confirmed by subprocess: false in skills.json.\n- [DATA_EXFILTRATION]: Outbound network access is disabled in the skills.json configuration, ensuring that no data can be sent to external servers.\n- [CREDENTIALS_UNSAFE]: Code examples and templates use clearly marked placeholders (e.g., YOUR_KEY, YOUR_API_KEY_HERE, token: 'your-api-token') for instructional purposes. No sensitive files are accessed and no hardcoded credentials were found.\n- [PROMPT_INJECTION]: The skill is configured to read all files in the repository (**/*) to perform analysis for documentation generation. This creates a surface for indirect prompt injection from the codebase content. Ingestion points: **/* (filesystem read permission). Boundary markers: Absent. Capability inventory: None (subprocess and outbound network are disabled). Sanitization: Absent. The risk is evaluated as SAFE due to the lack of any exploitable capabilities.

@tank/github-docs