The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's YAML frontmatter defines a lifecycle hook that executes a local script when the agent's task is stopped.\n
Evidence: node ${CLAUDE_PLUGIN_ROOT}/scripts/phase4-api-stop.js in the hooks section of SKILL.md.\n- [EXTERNAL_DOWNLOADS]: The skill provides instructions for the agent or user to add a remote Model Context Protocol (MCP) server, which involves fetching tool definitions from an external service.\n
Evidence: Command to add MCP server from https://api.bkend.ai/mcp in SKILL.md.\n- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface due to the lack of input validation and boundary markers when processing project files.\n
Ingestion points: The agent is instructed to read and write to files in the src/api/ and docs/ directories, which could contain untrusted data.\n
Boundary markers: Absent; there are no instructions to use delimiters or ignore embedded instructions when reading these files.\n
Capability inventory: The agent is granted access to high-privilege tools including Bash, Write, and Edit, which could be misused if malicious content is processed.\n
Sanitization: Absent; the skill contains no logic for sanitizing or validating the content of the files before they are used in agent operations.

phase-4-api