The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to scan the project environment, such as reading package.json, Makefile, and pyproject.toml. It also directs the agent to execute a generated initialization script using bash init.sh to verify project health.
[DYNAMIC_EXECUTION]: The skill generates a shell script (init.sh) at runtime based on the detected project structure and subsequently executes it. While this is a core part of its functionality for environment initialization, it involves runtime script generation and execution.
[INDIRECT_PROMPT_INJECTION]: The skill extracts startup and test commands from external project files to populate the init.sh script, creating a potential vector for indirect injection if these files are compromised or malicious.
Ingestion points: Reads project configuration files (package.json, Makefile, pyproject.toml) and the AGENTS.md file to determine environment setup steps.
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious content within the scanned files.
Capability inventory: Execution of generated shell scripts (bash init.sh), file system read/write access, and modifications to AGENTS.md and tasks.json.
Sanitization: The skill lacks automated sanitization or validation logic for the commands it extracts from project files before incorporating them into the executable init.sh script.

harness-step3-session-management