The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/improve.py utilizes subprocess.run() to interact with the system environment and external tools. Specifically, it executes git rev-parse --show-toplevel to locate the repository root and various gh (GitHub CLI) commands to check authentication status and list issues from the hardcoded repository nextwork-projects/ai-video-editor-cooked. Additionally, SKILL.md provides instructions for the agent to execute shell commands for logging changes, validating plugins, and closing GitHub issues.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by ingesting and processing data from untrusted or externally-influenced sources.
Ingestion points: scripts/improve.py reads data from ~/.ai-video-editor/suggestions.jsonl, edits/*/corrections.jsonl, */review.json, and open GitHub issues labelled 'feedback' via the GitHub CLI.
Boundary markers: No explicit delimiters or instructions are used to separate the ingested feedback content from the agent's primary instructions.
Capability inventory: The skill instructions direct the agent to modify repository files including PRINCIPLES.md, SKILL.md, source code (e.g., quality.py), and CHANGELOG.md. It also performs shell execution for testing and validation.
Sanitization: The script performs grouping and similarity analysis but does not sanitize the content of the suggestions or issue bodies for malicious prompt instructions.

improve