The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs management tasks using git merge, git commit, and git diff for file conflict resolution, and tmux list-panes and tmux kill-pane for process monitoring and recovery. These actions are consistent with the infrastructure purpose of the skill.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted data from other agents without adequate isolation.
Ingestion points: Reads agent-generated handoff summary blocks and task metadata checkpoints in the context-overflow and agent-crash-recovery modules.
Boundary markers: The skill does not use explicit delimiters or instructions to ignore embedded commands when processing agent outputs.
Capability inventory: The framework can modify Git repository state and terminate processes based on the content of ingested task and status data.
Sanitization: No input validation or filtering is performed on the summaries or metadata before they are used to guide subsequent recovery actions.

damage-control