The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external artifacts which presents a potential injection surface.
Ingestion points: Ingests content from SPEC.md, decisions.md, roadmap rows, and governing issue trackers.
Boundary markers: Explicitly instructs the agent to treat all ingested content as data, never as instructions, and to report any directives found within artifacts as findings.
Capability inventory: Performs command execution (git, node) and appends findings to local ledger files (progress.md, planning-findings.md).
Sanitization: Relies on logical segregation of data and instructions as its primary defense.
[COMMAND_EXECUTION]: The skill executes local shell commands and Node.js scripts as part of its artifact verification and snapshotting workflow.
Evidence: Invokes git rev-parse HEAD and node scripts/pre-execution-snapshot.mjs (both build and verify modes) to ensure artifact integrity.

review-spec