The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a local shell script scripts/landed.sh to search the git history and branch names for ticket IDs. While the script is bundled with the skill and uses safety practices like fixed-string matching and argument delimiters, it represents a controlled execution of shell commands based on tracker data.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from ticket bodies and comments to classify work and identify blockers. This data could contain instructions intended to manipulate the agent's classification logic or subsequent ticket updates.
Ingestion points: Step 1 in SKILL.md specifies reading every open ticket's body and all comments.
Boundary markers: The instructions do not define specific delimiters or warnings to ignore instructions embedded within the ticket content.
Capability inventory: The skill has the capability to update ticket statuses and post comments (Step 6) and execute the landed.sh script.
Sanitization: There is no mention of sanitizing or escaping the ticket content before the agent evaluates it for classification.

sweep-backlog