The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could contain malicious instructions aimed at influencing the agent's behavior.
Ingestion points: The skill ingests user-supplied goals, specs, and issue references from external hosted trackers in Stage 1 and Stage 3 (SKILL.md).
Boundary markers: The skill uses a structured Markdown ledger to maintain state and separate ticket data, but it does not specify explicit delimiters or warnings to ignore instructions embedded within the ticket bodies passed to subagents.
Capability inventory: The orchestrator dispatches subagents with the power to modify code via /vibe-implement and performs direct git operations including git worktree, git branch, and git commit across assigned workspaces.
Sanitization: While the instructions dictate that subagents should only implement specific tickets and the orchestrator should avoid reading full diffs, there is no automated sanitization or filtering of the external content ingested from trackers.

vibe-goal