The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to identify and terminate system processes and delete local directories to resolve Xcode build issues.
- Evidence: skills/xcode-debugging.md and SKILL.md contain commands for process inventory (pgrep -lx xcodebuild) and termination (kill -TERM <PID>). It also prescribes recursive deletion (rm -rf) of directories such as DerivedData, Pods/, and .build/.
- Risk: While these operations are targeted at the development environment, improper identification of process IDs or directory paths could lead to unintended disruption of other system processes or data loss.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data sources, which could be manipulated to influence agent behavior.
- Ingestion points: The agent is instructed to read crash logs (.ips, .crash files) as seen in SKILL.md and skills/lldb.md, and to parse dependency management files such as Package.swift, Package.resolved, and Podfile.lock in skills/build-debugging.md.
- Boundary markers: The instructions do not define clear delimiters or specific warnings to ignore embedded instructions within these external files.
- Capability inventory: The agent has the capability to execute shell commands (xcodebuild, simctl), terminate processes (kill), and modify the file system (rm).
- Sanitization: No sanitization or validation protocols are described for the external content before it is processed by the agent.

axiom-build