The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the firebase CLI tool to perform various project management tasks such as uploading symbols and managing Remote Config.
[COMMAND_EXECUTION]: The scripts/fcm-send.sh script uses curl, openssl, and python3 to construct and send authenticated HTTP requests to Google APIs.
[DATA_EXFILTRATION]: The skill accesses a sensitive credential file located at ~/.nutrisport/firebase/service-account.json. This access is necessary for the intended function of the skill (authenticating with Firebase) and is documented with instructions for secure local storage and .gitignore protection.
[EXTERNAL_DOWNLOADS]: The script interacts with well-known Google services at oauth2.googleapis.com and fcm.googleapis.com to obtain access tokens and send messages.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes user-provided or agent-retrieved data (notification titles, bodies, and data payloads) and sends it to an external API.
Ingestion points: CLI arguments passed to scripts/fcm-send.sh (e.g., --title, --body, --data).
Boundary markers: None present in the prompt interpolation.
Capability inventory: Network POST via curl in scripts/fcm-send.sh.
Sanitization: The script uses Python's json.dumps to correctly escape data before sending the JSON payload, which prevents schema confusion but does not filter for malicious instruction content.

firebase-ops