The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from bug reports and user-provided descriptions which are used to drive a high-privilege workflow involving source code edits and shell commands.
Ingestion points: The skill accepts a bug-id or description as its primary argument and reads existing bug files in SKILL.md (Phase 4 and Phase 6).
Boundary markers: The instructions do not specify delimiters or provide warnings to the agent to ignore instructions embedded within the bug descriptions.
Capability inventory: The skill has access to Write and Edit for modifying source code, Bash for executing repository commands (git), and the Agent tool to trigger other skill workflows.
Sanitization: There is no explicit sanitization or validation of the input bug data before it is processed by the agent.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform repository management tasks.
Evidence: In Phase 3, the skill executes git rev-parse --is-inside-work-tree to verify the environment and git checkout -b hotfix/[short-name] [base-ref] to manage branches. These actions are appropriately gated by the AskUserQuestion tool for user confirmation.

hotfix