The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing shell commands to interact with git repositories.
The instruction python3 scripts/scan_traces.py executes a local script that uses subprocess.run to call git log, git diff, and git diff --stat.
While the script uses list-based arguments for subprocess, the execution of shell tools based on repository content is a standard but noteworthy capability.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the ingestion of untrusted external data.
Ingestion points: The scripts/scan_traces.py script reads commit messages, PR bodies, and code diffs from the target repository.
Boundary markers: The skill lacks explicit boundary markers or instructions to the agent to disregard natural language instructions found within the git traces.
Capability inventory: The agent is granted the capability to generate content and perform distribution tasks such as publishing releases to GitHub, articles to Dev.to, and launching ad campaigns via tools.
Sanitization: There is no evidence of filtering or sanitization of the technical diffs or commit messages before they are processed by the LLM to 'Understand the product' and 'Rewrite to human language'.
[CREDENTIALS_SAFE]: The skill follows security best practices for secret management by instructing the user to provide API keys as environment variables or tool-specific configurations rather than hardcoding them.

releazer