The Agent Skills Directory

[SAFE]: The skill implements a set of read-only reconnaissance tools using standard git and gh (GitHub CLI) commands. It gathers repository metadata, issue details, and milestone status to recommend follow-up actions.
[COMMAND_EXECUTION]: The skill uses git commands (status, branch, log, diff) and gh CLI commands to inspect the environment. These are restricted to gathering situational awareness and do not perform destructive actions or unauthorized modifications without user approval for follow-up skills.
[DATA_EXFILTRATION]: While the skill accesses GitHub and Git logs, all network operations are directed to github.com via the official gh CLI or standard Git protocols. No non-whitelisted domains or evidence of sensitive credential harvesting were found.
[PROMPT_INJECTION]: The instructions do not contain attempts to bypass AI safety filters or override system prompts. It uses 'Plan Mode' as a structured workflow constraint to ensure user approval before any mutating actions are suggested.
[DYNAMIC_CONTEXT_INJECTION]: The skill does not use the !command`` syntax for load-time execution. It relies on explicit tool calls and standard agent interaction flows.

recon