The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute local shell scripts to manage the environment.
Evidence: bash aikey/scripts/list-skills.sh is used in Step 4 to check for duplicate skills before creation.
[DYNAMIC_EXECUTION]: The skill's primary function is to generate new directories, Markdown instructions, and executable shell scripts, then run them for validation.
Evidence: Step 4 defines a structure including scripts/ for 'repetitive and deterministic tasks'.
Evidence: Step 5 ('Run it once before delivery') explicitly requires the agent to execute the newly created skill and its associated scripts in a new context to verify performance.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted session data to generate persistent logic, which could formalize malicious instructions into a permanent skill.
Ingestion points: The entire 'current session process' is used as the primary source for skill extraction (Step 1).
Boundary markers: The skill includes 'Gatekeeping' (门禁) logic to check for task completion and verification, but lacks specific sanitization for prompt injection patterns within the extracted logic.
Capability inventory: The skill has the ability to write files to the filesystem and execute shell commands.
Sanitization: No explicit escaping or filtering is applied to the extracted session content before it is written into a new SKILL.md or script file.

aikey-skillify