The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill uses dynamic context markers in SKILL.md to execute shell commands at load time. These commands are limited to node --version, python3 --version, and curl --version. These are benign version checks intended to verify the local environment's readiness for debugging tasks.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external API sources and instructions from the Lindy dashboard, creating a potential surface for indirect injection if that data were maliciously crafted.
Ingestion points: Data is ingested from the Lindy API (api.lindy.ai) and manual inputs from the agent's task history (SKILL.md).
Boundary markers: There are no explicit boundary markers or delimiters defined for the agent to distinguish between data and instructions when processing these diagnostics.
Capability inventory: The skill possesses the Bash tool (used for curl and npx) and Write tool (used to save the support bundle in references/implementation-guide.md).
Sanitization: The skill provides manual sanitization guidance, instructing the user to remove credentials and personal data before sharing the bundle outside the workspace.
[EXTERNAL_DOWNLOADS]: The skill references the @lindy-ai/sdk package and uses npx to execute scripts. These resources are directly related to the vendor's purpose and the skill's primary function of debugging Lindy AI instances.
[DATA_EXFILTRATION]: The skill performs connectivity tests to well-known services including Slack, Google APIs, and Notion. These checks are used to verify integration health and do not involve sending sensitive user data to these or any unknown third-party domains.

lindy-debug-bundle