The Agent Skills Directory

[SAFE]: The skill's primary purpose is to automate the publication of documentation from local YAML files to a Confluence instance. All observed behaviors align with this purpose.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute acli (official Atlassian CLI) and curl. These tools are used appropriately to interact with the Confluence REST API and manage authentication.
[CREDENTIALS_UNSAFE]: The skill demonstrates safe credential handling by resolving the CONFLUENCE_API_TOKEN from environment variables or through a direct user prompt via AskUserQuestion, rather than hardcoding sensitive information.
[REMOTE_CODE_EXECUTION]: The skill executes a local Python script (pubcheck.py) located within its own assets directory. This script uses standard Python libraries to perform structural validation of the generated documentation before it is pushed to the server.
[PROMPT_INJECTION]: The skill employs a sub-agent for "fresh-eyes" verification. The prompt provided to this agent is strictly task-focused, providing necessary context and clear instructions for semantic validation without attempting to bypass safety guardrails or inject malicious instructions.
[DATA_EXFILTRATION]: Data transfer is limited to the user's configured Confluence site. The destination URL is retrieved directly from the user's acli configuration, minimizing the risk of unauthorized data redirection.

confluence-api-doc