The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses several shell commands to gather environmental context and manage the local documentation store.
Evidence: Uses dynamic context injection in SKILL.md (!`git rev-parse --abbrev-ref HEAD 2>/dev/null || true`) to resolve the current Git branch name at skill load time.
Evidence: Invokes the GitHub CLI (gh issue list) within the research phase to find related external context and issues.
Evidence: Uses standard shell utilities like mkdir -p to create documentation directories.
Evidence: Executes a local Python script (scripts/validate-frontmatter.py) to validate the integrity and parser-safety of generated YAML frontmatter.
[PROMPT_INJECTION]: The skill features an indirect prompt injection surface as it aggregates data from multiple sources to generate documentation.
Ingestion points: Command output (Git branch), conversation history, platform-specific memory files (MEMORY.md), and existing files in docs/solutions/.
Boundary markers: Instructs the agent to use labeled excerpt blocks and specific section headers to isolate external context from instructions.
Capability inventory: The skill can write files to docs/solutions/, update instruction files (AGENTS.md, CLAUDE.md) following user confirmation, and execute local shell/CLI commands.
Sanitization: Includes a dedicated Python validation script and specific YAML quoting rules to prevent data corruption and ensure correct parsing of untrusted content.

ce-compound