The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill handles sensitive information including Meta Access Tokens and App Secrets.
Evidence: Instructions guide users to store access_token, ad_account_id, app_id, and app_secret in a local file at .agents/meta-ads-credentials.json for persistent access.
Note: This represents standard local secret management practice. No hardcoded credentials were found in the skill content.
[COMMAND_EXECUTION]: The skill executes local Python scripts to perform its core functions.
Evidence: Commands such as python skills/meta-ads-connect/scripts/audit.py and python skills/meta-ads-connect/scripts/mutate.py are used to pull data and apply changes.
Note: These are internal skill scripts used for their intended purpose; no arbitrary command execution or shell injection vectors were identified.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes external data from the Meta Marketing API.
Ingestion points: Data is fetched by scripts/audit.py and stored in .agents/meta-ads-data.json before analysis.
Boundary markers: The instructions do not define explicit delimiters or 'ignore' instructions for the retrieved ad data (e.g., ad copy, campaign names).
Capability inventory: The agent can execute local scripts and write to the .agents/ directory.
Sanitization: There is no mention of sanitizing or filtering the ad data before it is parsed by the LLM.
[EXTERNAL_DOWNLOADS]: The skill references official Meta developer documentation and tools.
Evidence: Links to developers.facebook.com for token generation and API configuration.
Note: These are trusted domains for the skill's stated purpose.

meta-ads-connect