The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources and possesses the capability to act on that data without oversight. \n
Ingestion points: External GitHub PR comments and reviews fetched via gh pr view and gh api commands. \n
Boundary markers: None. The skill does not use delimiters or instructions to ignore potentially malicious commands embedded in review comments. \n
Capability inventory: The skill can execute shell commands (git, gh) and write to the local file system. \n
Sanitization: None. The skill is specifically instructed to fix 'critical bugs' immediately without asking based on feedback from these external sources. \n- [DATA_EXFILTRATION]: The skill references a hardcoded absolute file path on the local system: /Users/antonnovoselov/Documents/Vault/Projects/VivaDicta/what's new/whats-new-running.md. While used for logging, hardcoding personal paths exposes internal system structure and limits the skill to a specific user environment. \n- [COMMAND_EXECUTION]: The skill performs automated execution of git and gh (GitHub CLI) commands to push branches, create PRs, post comments, and merge code. \n- [PROMPT_INJECTION]: The skill interpolates user-supplied $ARGUMENTS directly into the system prompt instructions, which could be used to influence or override the intended workflow.

prcdx