The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill provides a mechanism for agents to store and retrieve unstructured text across sessions, creating a surface for malicious instructions to be persisted and later executed by an agent.
Ingestion points: Data enters the system via commands such as td log, td handoff, and td create as documented in SKILL.md and references/ai_agent_workflows.md.
Boundary markers: No documentation exists in SKILL.md regarding the use of delimiters or protective headers to distinguish between system state and user-generated content during retrieval.
Capability inventory: Agents are instructed in SKILL.md to use the output of td usage and td next to determine their current focus and next actions.
Sanitization: The instruction files do not describe any sanitization or validation processes for the data stored in the local .todos/ SQLite database.
No Code (SAFE): The skill consists entirely of documentation and does not include scripts, binaries, or instructions to install external software at runtime.

td-task-management