The Agent Skills Directory

[DATA_EXFILTRATION]: The skill uses the exarchos_orchestrate tool with the create_issue action to transmit diagnostic data, including event logs and state information, to external VCS providers (GitHub, GitLab, Azure DevOps). While this is the intended functionality for reporting bugs, it constitutes the transfer of internal session data to external platforms.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection as it processes untrusted data from session history and MCP event logs (Step 2 and Step 3) to generate its reports.
Ingestion points: Workflow event streams via exarchos_event query(stream) and conversation history containing tool error messages.
Boundary markers: Absent; there are no instructions to use specific delimiters or to treat the ingested log data as untrusted text that should not be followed as instruction.
Capability inventory: The skill utilizes exarchos_orchestrate to create external issues and exarchos_workflow to modify workflow states.
Sanitization: None; the skill directly analyzes and incorporates the content of failed tool calls into generated issue drafts.
Mitigation: The skill includes a 'human-in-the-loop' requirement in Step 6, explicitly instructing the agent to present draft issues to the user and wait for confirmation before calling the create_issue tool.

dogfood