port-herdr-integration
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONPERSISTENCE
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions for executing Go commands (
go run,go build,go test), project-specific tools (herdrsync), and specific agent reporting commands (agent report,agent report-session). These are part of the intended development and integration testing workflow. - [INDIRECT_PROMPT_INJECTION]: The skill describes a process that ingests external data from Herdr sync reports and upstream asset repositories.
- Ingestion points: The skill reads Herdr sync reports (
internal/agentactivity/manifests/report.md) and vendored upstream assets (internal/agentintegration/upstream/). - Boundary markers: The skill mentions using "fenced marker regions" in configuration files to delimit Sidecar-managed entries and ownership markers (
sidecar-integration:) to identify its own changes. - Capability inventory: The skill utilizes subprocess spawning for agent reporting, execution of Go toolchain commands, and direct file system modifications for installing hooks.
- Sanitization: The skill describes a trace capture process that sanitizes data "by construction," recording only event names and field keys while excluding sensitive values like prompt text, credentials, or environment values.
- [PERSISTENCE]: The skill describes a process for installing lifecycle hooks and plugin assets into the configuration files of other agents (e.g., Mastra Code, Kilo, Claude). This is the primary purpose of the integration skill and is managed through specific installer logic that uses marker tokens for identification and removal, ensuring Sidecar can monitor agent state across sessions.
Audit Metadata