The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs shell command execution to detect the presence of external LLM CLI tools (command -v codex, command -v gemini) and searches for them in system paths such as ~/.local/bin. If detected, these tools are executed to perform automated plan reviews. Additionally, the skill executes user-defined test commands and verification steps as part of the implementation workflow.
[DATA_EXFILTRATION]: The skill is designed to send technical implementation plans (plan.md) to external command-line interfaces (codex, gemini) for review purposes. While these are recognized developer tools, this behavior involves transmitting project metadata and architectural plans to external processes.
[PROMPT_INJECTION]: The skill serves as an orchestrator that ingests project documentation and source code to generate implementation specs and plans, creating an attack surface for indirect prompt injection.
Ingestion points: The skill reads local repository files including CLAUDE.md, AGENTS.md, and project source code during the exploration and specification phases.
Boundary markers: The instructions do not specify the use of delimiters or specific warnings to ignore instructions embedded within the codebase being analyzed.
Capability inventory: The skill possesses capabilities for file system read/write operations, shell command execution (tests), and invocation of external binaries.
Sanitization: There is no explicit evidence of content sanitization or validation performed on the repository data before it is processed by the agent.

feature