The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides structured command-line examples for the acpx utility, enabling capabilities like background monitoring, parallel task execution, and automated retries.\n- [PROMPT_INJECTION]: The 'Crash-Resilient Overnight' and 'Parallel Development' patterns instruct the agent to read and follow instructions from external files (e.g., tasks/plan-*.md and tasks/spec-*.md). This establishes an indirect prompt injection vulnerability surface where untrusted data could influence agent behavior.\n
Ingestion points: Local project files in the tasks/ directory are read using the Read and Glob tools as specified in the command examples in SKILL.md.\n
Boundary markers: Absent; the provided prompt templates do not use delimiters or instructions to prevent the agent from obeying commands embedded within the input files.\n
Capability inventory: The commands grant the agent access to powerful tools including Bash, Write, and Edit, which could be abused if the agent is compromised by malicious file content.\n
Sanitization: Absent; no validation or sanitization of the content from the implementation plans is performed.

acpx-patterns