The Agent Skills Directory

[SAFE]: The skill implements a legitimate documentation review workflow. It operates locally on repository files to identify implementation drift, stale examples, and clarity issues.\n- [SAFE]: No external network operations, downloads, or remote code execution patterns were detected. The included shell scripts (scripts/collect-code-files.sh, scripts/collect-doc-files.sh) use standard find commands to locate files by extension within the current directory.\n- [SAFE]: No obfuscation (Base64, zero-width characters, homoglyphs) or credential exposure (hardcoded API keys, tokens) was found in the skill's instructions or scripts.\n- [SAFE]: The skill does not attempt to escalate privileges or establish persistence. It follows a passive analysis role, only suggesting modifications if explicitly requested by the user.\n- [INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it is designed to ingest and analyze arbitrary documentation and source code files from a repository. However, the risk is evaluated as safe because the skill lacks capabilities for network communication or arbitrary code execution.\n
Ingestion points: The agent reads repository files identified by the find scripts in the scripts/ directory.\n
Boundary markers: No specific delimiters or "ignore instructions" warnings are defined in the SKILL.md for processing external file content.\n
Capability inventory: The skill does not perform network operations, subprocess executions (outside the initialization scripts), or dynamic code evaluation.\n
Sanitization: There is no explicit sanitization or filtering of the content read from files.

docs-drift-review