The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources such as ticket management systems (Jira, Linear, GitHub Issues), error trackers (Sentry, Bugsnag), and logging platforms.
Ingestion points: Phase 1 fetches ticket descriptions, acceptance criteria, stack traces, and log entries using MCP tools or user input.
Boundary markers: The skill lacks explicit delimiters to isolate external content from instructions. It relies on a summary step (Phase 1.5) and a manual plan approval step (Phase 4.3) as human-in-the-loop checks.
Capability inventory: The skill can execute shell commands (git, test runners, VCS CLIs), write to local configuration files (.supplement.md), and invoke other specialized skills.
Sanitization: There is no evidence of sanitization or escaping of the ingested external content before it is used to generate summaries or implementation plans.
[COMMAND_EXECUTION]: The workflow executes various shell commands based on configuration stored in the .supplement.md file.
Evidence: Phase 2, 5, 6, and 8 use variables like $MAIN_BRANCH, $BRANCH_PATTERN, $TEST_COMMAND, and $VCS_CLI to construct and run git commands and test suites. While these are intended to be user-provided environment settings, they represent a significant capability surface within the skill's workflow.

ticket-delivery