The Agent Skills Directory

[SAFE]: The skill implements a legitimate documentation co-authoring workflow. It guides the user through context gathering, brainstorming, drafting, and testing phases without attempting to bypass safety controls or exfiltrate data.
[DATA_EXPOSURE]: While the skill requests access to project context, team channels (Slack/Teams), and shared documents (Google Drive/SharePoint), it does so using standard platform integrations and explicit user consent. This is necessary for its primary purpose of writing context-aware documentation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external content such as info-dumps and shared documents. It mitigates potential instruction injection from these sources by including a 'Reader Testing' stage where the output is verified by a 'fresh' agent (sub-agent) with no prior context, ensuring the final document is clear and doesn't carry over malicious instructions.
[COMMAND_EXECUTION]: The skill uses standard file manipulation tools like create_file and str_replace to manage the document drafting process. These operations are scoped to the project directory and the specific document being authored.

doc-coauthoring