The Agent Skills Directory

[DATA_EXFILTRATION]: Accesses sensitive local files. The 'Framework & Config Explorer' agent (Agent 4) is explicitly instructed to search for .env and .env.example files during its discovery process.\n- [DATA_EXFILTRATION]: Reads local agent configuration files. The skill executes shell commands to inspect ~/.claude/mcp_config.json and ~/.claude.json to verify MCP server availability.\n- [COMMAND_EXECUTION]: Executes shell commands for environment discovery. The skill uses grep and file existence tests to identify project types and detect installed testing frameworks.\n- [COMMAND_EXECUTION]: Generates and suggests installation commands. It identifies missing dependencies and provides the user with shell commands to install them.\n- [COMMAND_EXECUTION]: Employs dynamic agent task generation. The skill dynamically constructs and spawns sub-agents using the Task() function with prompts generated from discovered codebase information.\n- [PROMPT_INJECTION]: Vulnerable to indirect prompt injection (Category 8).\n
Ingestion points: Discovery agents read various source files, directory structures, and configurations from the local project (SKILL.md, Phase 1).\n
Boundary markers: The skill uses simple markdown headers to delineate data but lacks explicit instructions for sub-agents to ignore potentially malicious instructions embedded in the codebase.\n
Capability inventory: The skill can spawn multiple sub-agents with executable tasks, write files to the local disk, and invoke other functional skills.\n
Sanitization: There is no evidence of sanitization or escaping applied to codebase content before it is processed by the specialist agents or the 'Doubt Agent'.

test-spec-gen