The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute multiple local shell scripts and tools, including .guardrails/scripts/find-items.sh, grill-requirements, analyze-risks, design-architecture, and .guardrails/scripts/new-id.sh. These commands are used to verify preconditions and mint identifiers during the planning process.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests untrusted data that is then used to influence the agent's output.
Ingestion points: The agent reads requirement items from the project ledger via the .guardrails/scripts/find-items.sh show ID command as specified in SKILL.md.
Boundary markers: There are no explicit instructions for the agent to use delimiters or ignore instructions potentially embedded within the requirement data.
Capability inventory: The agent has the capability to execute shell scripts (found in .guardrails/scripts/), write implementation plan files to docs/plans/, and call other skills like develop-change or merge-change.
Sanitization: The skill lacks explicit sanitization, escaping, or validation steps for the content retrieved from the requirements ledger before it is processed into the implementation plan.

plan-change