The Agent Skills Directory

[PERSISTENCE]: The skill's core purpose is to facilitate the creation of long-running, autonomous background agents described as 'daemons,' 'scheduled workers,' and 'persistent background agents.' The 'Deliverables' section explicitly requires the design of 'daemon, cron, or persistent worker' mechanisms, which are standard vectors for establishing persistence.
[DYNAMIC_EXECUTION]: The procedure instructs the agent to 'Return a runnable spec with lifecycle hooks' and design logic for 'idempotence, observability, [and] safe retries.' This involves generating executable task logic or scripts meant to run repeatedly in the background without user oversight.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a surface for indirect injection because it processes a user-provided 'mandate' or 'target' to generate persistent automation. If the instructions for the worker's task are derived from untrusted sources (e.g., summarizing an external document or processing API data), malicious instructions could be embedded into the long-running process.
Ingestion points: SKILL.md Procedure steps 1 and 2 ('target', 'success condition', and 'worker's single mandate')
Boundary markers: Absent. The procedure does not suggest using delimiters or warnings to ignore embedded instructions in the mandate data.
Capability inventory: Generation of daemons, cron jobs, and persistent background workers (SKILL.md Deliverables).
Sanitization: Absent. While it mentions 'no-touch boundaries,' there is no technical validation or escaping required for the data used to define the worker's logic.
[COMMAND_EXECUTION]: The skill encourages the design of tasks that execute recurring actions. Although the skill itself is a high-level design template, the resulting 'runnable specs' are intended to perform system-level operations automatically.

mordenkainen-s-sword