The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were detected. The skill follows documented procedures for task management and planning.
[DATA_EXPOSURE]: The skill accesses Jira issue data (via MCP) and local session files in the .oh/ directory. This data is used to generate subtasks and is not exfiltrated to unauthorized external domains.
[COMMAND_EXECUTION]: The skill uses codebase exploration tools and mentions bash as a fallback mechanism for searching files. This is standard behavior for development-oriented agent skills to understand project architecture.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it processes untrusted data from Jira issues and codebase files.
Ingestion points: Reads data from .oh/<session>.md files, Jira issue descriptions via jira_get_issue, and file content during codebase exploration.
Boundary markers: Uses markdown headers (e.g., ## Solution Space) to delimit data in session files, but lacks explicit boundary markers when interpolating Jira issue content or codebase snippets into subtask descriptions.
Capability inventory: Includes filesystem read/write, Jira issue creation/linking via MCP tools, and codebase search capabilities.
Sanitization: No explicit sanitization or filtering of the ingested external content is mentioned before it is formatted into new Jira subtasks.

jira-plan