The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in context-essentials.md mandate the execution of bash .ai_state/init.sh during implementation and review stages. This pattern executes a shell script residing within the project's local state directory.
[EXTERNAL_DOWNLOADS]: In the System workflow route defined in SKILL.md, the skill references npx ecc-agentshield scan, which involves downloading and executing a package from the NPM registry.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it processes various project files to manage its state and hand off tasks between models. Maliciously crafted content in these files could influence agent behavior.
Ingestion points: The agent reads state and configuration from .ai_state/project.json, .ai_state/progress.md, .ai_state/tasks.md, and handoff.md.
Boundary markers: There are no explicit instructions or delimiters used to ensure the agent disregards potential instructions embedded within the data read from these files.
Capability inventory: The agent has extensive capabilities, including executing shell commands, running tests, and managing Git operations.
Sanitization: No evidence of input sanitization or validation is present for the project-level data ingested during the workflow.

pace