The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill framework ingests untrusted data from Jira, such as Acceptance Criteria, Story descriptions, and issue comments, which are synchronized to local markdown files. This external content is used to drive test planning and automation tasks, representing an indirect prompt injection attack surface.
Ingestion points: references/acli-integration.md details the synchronization of Jira issue content into .context/PBI/ markdown files.
Boundary markers: The provided references do not specify explicit delimiters or 'ignore' instructions for the ingested Jira content.
Capability inventory: The framework is designed to execute shell commands (bun, gh), perform network operations (curl), and write to the filesystem.
Sanitization: The skill includes doctrine for converting Markdown to Atlassian Document Format (ADF) in references/jira-publishing-gotchas.md to ensure valid data transmission to Jira.
[DYNAMIC_CONTEXT_INJECTION]: The skill documentation mentions the !command`` syntax, suggesting it to users for interactive tasks like authentication.
Evidence: In references/preflight-gate.md, the skill suggests that a user can run !gh auth login to authenticate the GitHub CLI.
[EXTERNAL_DOWNLOADS]: The skill references and facilitates the installation of tools and API integrations from well-known technology services.
Evidence: references/preflight-gate.md describes the installation of Playwright browsers via bun run pw:install.
Evidence: The framework integrates with established services like Tavily and Resend for search and email capabilities as noted in references/preflight-gate.md.

agentic-qa-core