The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill ingests untrusted data from the Instantly API and interpolates it into terminal logs and Slack alerts without sanitization or delimiters.
Ingestion points: client.campaigns.list and client.accounts.list calls in SKILL.md retrieve campaign names and email addresses.
Boundary markers: Absent; data is directly embedded into strings like *${c.check}*: ${c.message}.
Capability inventory: The skill uses fetch for network operations and console.log for output, providing a path for instructions to influence agent behavior or be sent to external channels.
Sanitization: No escaping or validation is performed on strings retrieved from the API before they are processed.

instantly-observability