The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides an 'Autonomy contract' that explicitly overrides the agent's standard behavior for seeking user confirmation. It instructs the agent to 'proceed rather than asking permission' and claims 'questions cannot be answered mid-run', which removes important interactive safety checkpoints.
[PROMPT_INJECTION]: The 'Turn-ending check' section enforces a continuation policy ('the turn is not over. Execute, then end') that discourages the agent from stopping for feedback, even when it might be appropriate to do so.
[PROMPT_INJECTION]: The skill includes instructions to disregard platform-level context limitations ('do not stop, trim your work, or propose a new session on account of it'), which attempts to override system-imposed operational boundaries.

autonomous-continuation