The Agent Skills Directory

[PROMPT_INJECTION]: The skill demonstrates a potential surface for indirect prompt injection as it ingests data from external trackers (ticket IDs, dependency links, and engine labels) and interpolates them into instructions for sub-agents. While no malicious intent is present, the architectural pattern relies on the integrity of the external tracker data.
Ingestion points: Issue tracker APIs (JIRA, Linear, Azure DevOps) and local markdown files.
Boundary markers: The sub-agent prompt template lacks explicit delimiters or 'ignore embedded instructions' warnings for interpolated ticket metadata.
Capability inventory: The skill spawns sub-agents that invoke task-executor, which possesses capabilities for file system modification, git operations, and code execution via the Codex engine.
Sanitization: The skill does not explicitly describe sanitization or validation logic for data fetched from issue trackers before it is used in sub-agent prompts.
[COMMAND_EXECUTION]: The skill performs shell-level operations for repository management, including git worktree creation, git merge, and git rebase. These operations are well-scoped to the skill's purpose of orchestrating code changes across multiple tickets and include safety checks for merge conflicts.

epic-executor