The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill documents an installation method in SKILL.md that pipes a remote script directly into the shell: curl -fsSL https://raw.githubusercontent.com/yasserstudio/gpc/main/scripts/install.sh | bash. While the source belongs to the vendor (yasserstudio), this pattern is inherently risky as it executes unverified remote code.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download and installation of external software from multiple sources:
The @gpc-cli/cli package from the npm registry.
The yasserstudio/gpc-action from GitHub.
Standalone binaries from the author's GitHub repository.
[COMMAND_EXECUTION]: The environment detection script scripts/detect_gpc.mjs uses execSync to run shell commands (e.g., gpc --version, gpc auth status). These commands are hardcoded to check for the presence and configuration of the GPC tool.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it ingests and processes untrusted data that could contain malicious instructions:
Ingestion points: Reads git commit history via gpc changelog generate, Play Store crash/ANR data via gpc vitals, and user reviews via gpc reviews list.
Boundary markers: The instructions recommend using structured output formats (JSON/CSV) which provide structural isolation, but they do not define explicit LLM boundary delimiters for the processed content.
Capability inventory: The agent has the capability to perform sensitive operations based on processed data, such as gpc releases promote and gpc releases upload.
Sanitization: The skill mentions a --strict flag for changelog generation to detect jargon, but does not implement comprehensive sanitization for all external data streams.

gpc-ci-integration