The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from project documentation, creating a surface for indirect prompt injection where malicious instructions embedded in these files could attempt to manipulate agent behavior.
Ingestion points: The agent is instructed to read docs/BRIEF.md, docs/GLOSSARY.md, and docs/STORIES.md to gather context for the UX review.
Boundary markers: The skill uses a specific UX: block for its output in docs/STORIES.md, but it lacks explicit boundary markers or instructions to ignore potential commands within the input documentation files.
Capability inventory: The skill possesses capabilities to execute shell scripts (./build.ps1, tools/snap.ps1, tools/grid-lint.ps1) and perform file writes to documentation.
Sanitization: No sanitization, escaping, or validation of the input documentation content is specified.
[COMMAND_EXECUTION]: The instructions direct the agent to run local shell scripts, specifically ./build.ps1 for building the app and tools/snap.ps1 for reviewing implementation snapshots. These commands are consistent with the skill's purpose in a development environment but constitute executable capabilities triggered by the agent.

ux-design