The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard system commands including git diff, find, and sed to identify modified files and extract metadata. It also provides an optional pre-commit hook script for local automation, which is a standard developer practice and does not involve malicious persistence.
[DATA_EXFILTRATION]: Data operations are strictly confined to the local filesystem (reading from .claude/ and writing to docs/workflow/) and an internal PostgreSQL database via MCP tools. No network activity to external or untrusted domains was identified.
[PROMPT_INJECTION]: No direct prompt injection patterns, safety bypasses, or instructions to override agent behavior were detected in the skill documentation.
[PROMPT_INJECTION]: Indirect Prompt Injection Analysis
Ingestion points: The skill reads content from local .agent.md and .md files and queries metadata from a PostgreSQL database.
Boundary markers: Metadata is parsed using YAML delimiters (---) and specific filename patterns.
Capability inventory: The execution environment utilizes Read, Write, Edit, and Bash tools to update documentation files and generate reports.
Sanitization: The process includes YAML frontmatter validation and Mermaid diagram syntax checking to ensure the integrity of the processed data.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute code from remote sources. It relies on local scripts and predefined MCP tools for its operations.

doc-sync