The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to inspect and modify external skill files, creating a surface for instructions embedded in those files to influence agent behavior.
Ingestion points: The agent reads and processes target skill directories, including SKILL.md and various files in references/ and examples/ folders as defined in the default workflow.
Boundary markers: Absent. The instructions do not direct the agent to wrap external content in delimiters or include warnings to ignore embedded instructions.
Capability inventory: The skill performs file reads (via rg and find), file writes for skill creation/repair, and executes local scripts (init_skill.py, package_skill.py).
Sanitization: Absent. No validation or filtering is specified for the content retrieved from external files before it is processed or written.
[COMMAND_EXECUTION]: The skill utilizes local shell utilities and scripts to perform environment checks and packaging.
Evidence: references/integrity-checks.md provides commands for rg (ripgrep) and find to verify file existence and remove caches. references/skill-creator-original.md provides execution patterns for scripts/init_skill.py and scripts/package_skill.py for initialization and distribution tasks.

skill-development