The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains deceptive metadata and licensing information that misrepresents its origin. Both SKILL.md and LICENSE.txt claim authorship and copyright by "OpenAI", a trusted organization, while the actual author is "tristanmanchester". This impersonation of a trusted vendor could lead to a misjudgment of the skill's safety and provenance.
[DATA_EXFILTRATION]: The scripts/meta_ads.py script includes functionality to read arbitrary local files via the --params-file flag or the @file syntax in the --set argument. These files can be uploaded to Meta's servers using the upload command or sent as parameters in API requests. This creates a surface for data exposure and exfiltration of sensitive local information if an agent is manipulated into reading system or credential files.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes untrusted data from the Meta Marketing API.
Ingestion points: External data is ingested from the Meta Graph API via the insights, list, get, and request subcommands in scripts/meta_ads.py.
Boundary markers: No explicit delimiters or instructions are used within the script's output to separate API-returned content from agent instructions.
Capability inventory: The skill possesses the capability to perform network mutations (POST/PUT requests) to the Meta API and read/write local files (via scripts/meta_ads.py).
Sanitization: There is no evidence of sanitization or filtering of the content received from the external API before it is presented to the agent.

meta-ads-control