The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of specialized CLI tools (rossi and eventb-animate) to validate, build, and model-check formal specifications. These commands are executed in the local shell environment.
Evidence: Found in SKILL.md and project README.md files, which provide command templates for the agent to run (e.g., rossi validate ., eventb-animate --json report.json model.zip).
[INDIRECT_PROMPT_INJECTION]: The skill operates on user-provided requirements and formal model files, creating a potential surface where malicious instructions could be embedded in the data processed by the agent.
Ingestion points: The agent reads and processes .eventb specification files (e.g., examples/counter/counter_abstract.eventb) and descriptive README.md files containing requirements.
Boundary markers: The skill relies on the strict formal grammar of the Event-B language to delimit data; no specific natural language boundary markers or 'ignore' instructions are defined for the requirements phase.
Capability inventory: The agent has the capability to execute shell commands via rossi (for formatting, validation, and building) and eventb-animate (for model checking, animation, and test generation) as defined in SKILL.md and references/tooling.md.
Sanitization: Inputs are subjected to strict semantic and syntactic validation by the rossi parser during the validate and build steps before any animation or model checking occurs.

eventb