The Agent Skills Directory

[COMMAND_EXECUTION]: The skill operates by executing the Meta Ads CLI through a Python safety wrapper located at scripts/meta_ads_agent.py. This script uses subprocess.run to manage command execution and risk classification.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from Meta Ads campaign names, creative content, and insight reports. This data could potentially contain malicious instructions intended to manipulate the agent's behavior.
Ingestion points: Commands like meta ads campaign list and meta ads insights get in scripts/meta_ads_agent.py retrieve data that is then processed by the agent.
Boundary markers: The skill instructions do not explicitly require the use of data delimiters or safety instructions when handling CLI output.
Capability inventory: The skill facilitates spend-modifying commands (create, update, budget) and account-modifying actions via the meta CLI.
Sanitization: While the safety script redacts authentication tokens, it does not sanitize descriptive text fields in the returned data for potential prompt injection payloads.
[METADATA_POISONING]: The SKILL.md file and LICENSE.txt claim the author is "OpenAI," which conflicts with the actual skill author "tristanmanchester." This misattribution is deceptive regarding the skill's origin.
[EXTERNAL_DOWNLOADS]: The skill documentation recommends installing the meta-ads package from PyPI, which is the standard mechanism for acquiring the Meta Ads CLI tool.

meta-ads-cli