The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs standard version control operations (e.g., git branch, git branch --show-current) to manage development tasks and verify branch state. These commands are restricted to the local environment and aligned with the skill's intended functionality as a developer assistant.
[INDIRECT_PROMPT_INJECTION]: The skill has an ingestion surface for untrusted data, as it reads ticket descriptions and developer comments from external issue trackers (such as Jira, GitHub, or Linear). This is a inherent risk for tools processing external documentation, but it is mitigated by the skill's structured Q&A loop and explicit user confirmation steps.
Ingestion points: External ticket bodies and comments (SKILL.md Step 1).
Boundary markers: None explicitly defined in the provided handoff templates.
Capability inventory: Local file writing (plans/ directory), git branch management, and invocation of planning/audit sub-agents.
Sanitization: External content is processed directly to extract architectural context without explicit sanitization routines.
[DATA_EXPOSURE]: The skill reads project configuration files (e.g., CLAUDE.md, AGENTS.md) and roadmap documents to ground its planning. This is standard behavior for an agent understanding a codebase and does not involve exfiltration of sensitive credentials or private keys.

synthesis-preplan