The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external API responses, which presents an attack surface where malicious content from a compromised API could influence the agent's context.
Ingestion points: Data is retrieved from https://api.salesloft.com/v2/me.json and other SalesLoft endpoints using curl commands in SKILL.md.
Boundary markers: The instructions do not define explicit boundary markers or delimiters for the data returned from the API calls.
Capability inventory: The skill utilizes the Bash tool to execute shell commands, read local files, and create compressed archives (tar).
Sanitization: The skill uses jq to extract specific JSON fields and sed to redact environment variable values.
[DATA_EXFILTRATION]: The skill accesses the local .env file to include configuration state in the debug bundle.
Evidence: The script executing cat .env 2>/dev/null | sed 's/=.*/=***REDACTED***/' accesses a sensitive file path. However, the use of sed to replace all values with a redaction string effectively mitigates the risk of credential exposure.

salesloft-debug-bundle