The Agent Skills Directory

COMMAND_EXECUTION (SAFE): The skill includes several shell scripts (context-recovery.sh, phase-tracker.sh, validate-phase.sh) that use standard Unix tools like grep, sed, and cat. These scripts facilitate project state management and do not perform network operations, access sensitive user data, or attempt privilege escalation.
PROMPT_INJECTION (LOW): The skill is susceptible to indirect prompt injection due to its automated context recovery mechanism.
Ingestion points: scripts/context-recovery.sh reads and echoes data from docs/MASTER_PLAN.md, docs/.phase-state.json, and phase-specific IMPLEMENTATION.md or TEST_PLAN.md files back to the agent.
Boundary markers: Absent; untrusted project content is interpolated into the agent's context without delimiters or explicit instructions to ignore embedded directives.
Capability inventory: The skill possesses the ability to read and write files within the project directory and execute local bash scripts; it does not have network egress or high-level system permissions.
Sanitization: Content read from the project documentation is not sanitized or escaped before being presented to the agent.

solo-builder