The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The script create_agent_memories_simple.py uses the uvx tool to download and execute the mcp-neo4j-memory package from the public PyPI registry to manage agent context in a Neo4j database.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute validation logic and interact with CLI-based testing tools, which is consistent with its role in managing system-level agent configurations.
[PROMPT_INJECTION]: The skill processes markdown-based agent definitions, presenting an indirect prompt injection surface.
Ingestion points: Metadata and instructions are ingested from markdown files located in .claude/agents/ as seen in the extract_agent_metadata logic in scripts/agent_detector_example.py.
Boundary markers: The system uses YAML frontmatter delimiters (---) to distinguish between configuration parameters and system prompt instructions.
Capability inventory: Agents managed by this skill can be configured with powerful tools such as Bash, Write, and Edit, as documented in the configuration reference.
Sanitization: The implementation uses yaml.safe_load() to ensure that configuration parsing does not trigger unsafe code execution through YAML deserialization.

manage-agents