The Agent Skills Directory

[COMMAND_EXECUTION]: Uses standard command-line tools such as curl, jq, and cat to manage ad campaign metadata and fetch performance metrics. These operations are restricted to interactions with the official Meta API.
[DATA_EXFILTRATION]: Network operations are directed exclusively to graph.facebook.com, which is a well-known technology service. This communication is required for the skill's stated purpose of managing ads.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it processes untrusted data from API responses. Evidence Chain: 1. Ingestion points: API responses from graph.facebook.com in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: shell commands including curl, jq, and cat (writing to /tmp) across SKILL.md. 4. Sanitization: Absent. This is a common structural risk for skills that ingest external data and does not represent a malicious finding.

meta-ads