The Agent Skills Directory

[COMMAND_EXECUTION]: The skill grants the subagent (Claude Code) full filesystem access and the ability to execute arbitrary commands via the Bash tool. The documentation explicitly encourages using these capabilities for complex system-level tasks like refactoring and scaffolding.
[COMMAND_EXECUTION]: The skill provides and describes a --yolo flag, which is documented as 'dangerously-skip-permissions'. This flag explicitly disables standard security safeguards (permission prompts), allowing for fully autonomous work that bypasses human-in-the-loop verification for sensitive operations.
[PROMPT_INJECTION]: The skill exhibits a significant indirect prompt injection surface.
Ingestion points: The skill takes a user-provided task description as input for the alma coding-agent run command.
Boundary markers: There are no visible delimiters or 'ignore' instructions to prevent the subagent from following malicious instructions embedded within the task string.
Capability inventory: The subagent has 'full filesystem access', shell access via Bash, and can run autonomously via the --yolo flag.
Sanitization: No evidence of input validation or sanitization is present. An attacker could provide a task that includes commands to exfiltrate files or install persistent backdoors, which the subagent might execute autonomously.

coding-agent