The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes potentially untrusted content from a local log file that could contain indirect prompt injection instructions.
Ingestion points: The file .qa-learnings/ledger.md is read and parsed at runtime.
Boundary markers: No explicit delimiters or boundary instructions are present in the skill to isolate the ledger content.
Capability inventory: The skill can create GitHub issues, labels, forks, and pull requests via the gh tool.
Sanitization: Content is previewed by the user in the terminal, requiring explicit approval before submission.
[DATA_EXFILTRATION]: The skill transmits local QA observations to the author's GitHub repository.
Evidence: Employs gh issue create and gh pr create to send data to the neonwatty/qa-skills repository.
Note: Data transfer requires explicit user selection and approval of the drafted content.
[COMMAND_EXECUTION]: Uses the Bash tool to execute various system and CLI commands.
Evidence: Executes gh commands for repository management and standard file system operations.
Note: These commands are integral to the stated workflow of the skill.

submit-learnings