The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill requires and accesses a sensitive GITHUB_PERSONAL_ACCESS_TOKEN stored in the agent's secrets to perform repository synchronization tasks.
[COMMAND_EXECUTION]: Executes shell commands via curl to verify GitHub credentials and to register new automations with the platform's internal API.
[DYNAMIC_EXECUTION]: Uses a piped shell command to execute a Python snippet (curl ... | python3 -c ...) for parsing JSON responses from the GitHub API during token verification.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by interpolating user-provided text into a prompt for a downstream automation.
Ingestion points: The $LOCAL_CHANGES variable, which accepts arbitrary user input in SKILL.md.
Boundary markers: Absent; the variable is directly inserted into the JSON payload of a curl command without delimiters or warnings to the downstream LLM.
Capability inventory: The skill facilitates the creation of persistent cron automations that have the capability to execute git commands and project build/test scripts.
Sanitization: Absent; the shell snippet performs simple string interpolation without escaping or validating the content of the $LOCAL_CHANGES variable.

upstream-fork-sync