The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests requirements from project-specific documents such as PRDs, epics, and architecture guides (Ingestion points: SKILL.md Steps 2.1 and 3.2). Boundary markers are present in the form of required source citations ([Source: ...]) and markdown comments (). The capability inventory includes file writes, git operations, and GitHub/Jira CLI interactions. Sanitization is implemented via automated redaction in defer-mutation.js and strict subset parsing in read-config.sh and yaml-subset.js.
[COMMAND_EXECUTION]: Executes local utilities and CLI tools including git, gh, and node to manage documentation and synchronize with issue trackers. These operations are restricted to documented project management tasks and often require user confirmation.
[DATA_EXFILTRATION]: Features a robust redaction mechanism in defer-mutation.js designed to scrub sensitive tokens and passwords from logs and tracker comments. It correctly manages Atlassian and GitHub credentials through environment variables rather than hardcoded secrets.
[DYNAMIC_EXECUTION]: Employs Python and Node.js for tasks such as hierarchical document ID calculation and configuration parsing. The execution logic is local to the skill and includes security hardening measures like sys.path isolation in Python to prevent local module hijacking.

create-story