The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains an explicit instruction to override environment security controls by stating that the "sandbox" must be turned off to allow the GitHub CLI to access configuration files like ~/.config/gh/hosts.yml, which typically contains authentication tokens.- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from pull request metadata and file paths retrieved via GitHub CLI commands. This data is used in shell script pipelines without boundary markers or sanitization, creating a surface for indirect injection if a repository contains maliciously named files or metadata.
Ingestion points: Pull request metadata and file lists retrieved via gh pr view in SKILL.md.
Boundary markers: None provided for external data used in shell commands.
Capability inventory: Shell execution via the Bash tool (including gh, shasum, base64, and sed), and PR modification via gh pr edit.
Sanitization: None identified in the provided shell snippets for handling potentially malicious file paths or PR content.- [COMMAND_EXECUTION]: The skill relies on extensive shell pipelines involving the GitHub CLI, shasum for computing diff anchors, and base64 for decoding file contents to automate PR management tasks.

tomtom-maps-sdk-js-pr-description