The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple CLI tools to perform its tasks. It utilizes a helper script scripts/select-mode.sh which invokes git, jq, and herdr to manage agent sessions and repository state. It directly invokes the jira CLI to verify issue existence and query status. It uses forge-specific tools (glab or gh) for merge request operations.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external Jira tickets, which serves as a potential vector for indirect prompt injection. 1. Ingestion points: Data is ingested in SKILL.md (Step 2 and the Local mode workflow) where ticket content is read and used to generate plans. 2. Boundary markers: There are no explicit boundary markers or instructions to ignore embedded commands within the ticket data described in the process. 3. Capability inventory: The skill has significant capabilities including file system modification (worktrees), source control operations (commit, push, MR), and agent spawning. 4. Sanitization: While the ticket_id is sanitized for session handles, the full ticket content is used for planning without detailed sanitization steps.
[DYNAMIC_EXECUTION]: The skill generates and then executes logic based on processed data. It produces an implementation plan using /plan-ticket and then executes that plan, potentially creating a new workflow. The workflow includes "env bootstrapping" and "dependency install", which involves running scripts and installing packages defined in the target project's repository.
[EXTERNAL_DOWNLOADS]: The implementation process involves "dependency install", which fetches code from external package registries based on the project requirements.

implement-ticket