The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various system-level commands to manage the optimization lifecycle:
Utilizes git commands (add, commit, revert, checkout, branch) to maintain a versioned history of skill improvements and to implement a "ratchet" mechanism that rolls back quality regressions.
Runs grep as part of its "Runtime Neutrality" gate to identify and remove strings tied to specific agent platforms (e.g., "Claude Code").
Executes node to run a local script (scripts/screenshot.mjs) and suggests running npx playwright to generate visual achievement cards.
The screenshot utility script uses execSync to call the open command for displaying results to the user.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation and use of external software and dependencies:
Recommends the installation of skillopt via pip, referencing academic work from Microsoft Research for validation-gated optimization.
Uses npx skills add for installation and npx playwright for utility functions, which involves fetching code from remote registries.
[PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection (Category 8) due to its core function of processing untrusted data:
Ingestion points: The skill reads and evaluates the content of SKILL.md files located in the local environment, which may be authored by third parties.
Boundary markers: The instructions do not specify the use of strict delimiters or isolation techniques when presenting target skill content to sub-agents for evaluation.
Capability inventory: The skill has significant local capabilities, including modifying files via git, executing shell commands, and spawning independent agent instances.
Sanitization: There is no mention of sanitizing or filtering input files before they are parsed by the agent or evaluated by the "judge" agents.

darwin-skill