The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the ETTrace runner CLI using Homebrew from emergetools/homebrew-tap/ettrace and clones the source code repository from https://github.com/EmergeTools/ETTrace to build the simulator framework.
[COMMAND_EXECUTION]: Performs extensive local command execution using xcodebuild for framework compilation, dwarfdump for UUID verification, and shell scripts for environment setup and dSYM management.
[PROMPT_INJECTION]: Identifies an indirect prompt injection surface in scripts/analyze_flamegraph_json.py. The script processes output_*.json files and outputs frame names directly to the summary.
Ingestion points: Reads trace data from output_*.json files in the RUN_DIR (referenced in SKILL.md).
Boundary markers: No delimiters or instructions are used to distinguish trace data from agent instructions in the output summary.
Capability inventory: Across SKILL.md and referenced scripts, capabilities include file system writes, shell command execution (brew, git, xcodebuild, cp, rm), and running local Python and Bash scripts.
Sanitization: There is no sanitization or escaping of the symbol names extracted from the JSON before they are printed to the summary, allowing potentially malicious symbol names to enter the agent's context.

ios-ettrace-performance