The Agent Skills Directory

[SAFE]: The skill's primary function is to enhance the safety and reliability of coding agents by requiring a discovery and planning phase before any code changes are made. It does not perform network requests, exfiltrate data, or attempt to bypass security filters.\n- [COMMAND_EXECUTION]: The skill includes a local Python script (scripts/validate_clarification_summary.py) to ensure the generated planning artifact follows the required template. This script is benign, using only standard Python libraries (argparse, pathlib, re, sys) to perform local text validation.\n- [PROMPT_INJECTION]: The skill manages a surface for indirect prompt injection as it requires the agent to read and process content from a user-provided repository. 1. Ingestion points: The agent reads codebase files, documentation, and tests (SKILL.md, Step 2). 2. Boundary markers: The agent is instructed to use a specific Markdown template (output-template.md) which uses clear structural headers to separate sections. 3. Capability inventory: Across its instructions and scripts, the skill can read local files, write a summary file to disk, and execute a local Python validator script. 4. Sanitization: While the skill lacks explicit semantic sanitization for malicious instructions within files, the structured workflow and external validator script enforce a process of technical grounding that mitigates the risk of the agent following embedded adversarial commands.

task-clarification-harness