The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill performs dynamic command execution using the eval command on strings provided by subskills.
Evidence: eval "<tool.check>" in Phase 2 of SKILL.md directly executes string content from the subskill frontmatter.
Evidence: The skill dynamically resolves and executes shell scripts and command sequences defined in subskill variables (Phase 3 and Phase 4).
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests procedural logic from subskill files and executes it as shell commands.
Ingestion points: Subskill YAML frontmatter fields such as check_commands, install_commands, and verify_commands processed in SKILL.md.
Boundary markers: Absent; there are no delimiters or warnings to ignore embedded instructions within the ingested data.
Capability inventory: Full shell command execution via bash and script execution capability.
Sanitization: Absent; the skill does not validate or sanitize the content of the commands before execution.

migration-wizard