The Agent Skills Directory

[COMMAND_EXECUTION]: The Denji CLI tool documentation describes a hooks feature within the denji.json configuration file. This allows for the execution of arbitrary shell commands during lifecycle events such as postAdd and postRemove. While primarily intended for automation (e.g., running formatters), it provides a mechanism for command execution defined in a project-local file.\n
Evidence: Found in SKILL.md and references/configuration.md. The documentation explicitly lists examples such as npx biome check --write and npx prettier --write as hook commands.\n- [DYNAMIC_EXECUTION]: The skill enables the dynamic execution of shell scripts or commands defined in the denji.json configuration file at runtime whenever the CLI performs specified actions.\n
Evidence: The hooks property in the denji.json schema is detailed in references/configuration.md as an array of commands executed by the CLI.\n- [INDIRECT_PROMPT_INJECTION]: The denji import command allows the ingestion of icon data from external JSON or text files, creating a vulnerability surface where maliciously crafted data files could attempt to influence the agent's behavior if processed unsafely.\n
Ingestion points: denji import reads manifest data from files or stdin as described in SKILL.md.\n
Boundary markers: The documentation does not specify the use of delimiters or warnings to isolate imported data from the agent's instructions.\n
Capability inventory: The skill has the capability to execute shell commands via its hooks system and modify project files.\n
Sanitization: No explicit sanitization or validation logic for the content of imported files is documented.

denji