uigraph
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data by reading various repository files (source code, READMEs, manifests) to derive system architecture and generate diagrams, schemas, and API specifications.
- Ingestion points: Artifact planning involves reading all project source files, READMEs, and main manifests (e.g.,
package.json,go.mod) as described inreferences/confirmation-workflow.md. - Boundary markers: The skill does not define explicit boundary markers for the data it interpolates, but it relies on strict schema-based generation rules.
- Capability inventory: The skill executes shell commands via
npx uigraph,go install, anduigraph-cli syncto manage the lifecycle of UIGraph artifacts. - Sanitization: Artifacts are validated post-generation using
uigraph sync --dry-runand a detailed set of manual verification rules provided inreferences/validation-rules.mdto ensure structural integrity. - [DATA_EXPOSURE]: The skill incorporates a robust safety mechanism in
references/confirmation-workflow.mdthat explicitly lists directories and files the agent must never read, including.envfiles, SSH keys, private certificates, and credentials. This proactively prevents accidental or intentional exfiltration of sensitive repository data. - [REMOTE_CODE_EXECUTION]: The skill instructs the agent to install and run its own CLI tools using
npx --yes uigraphandgo install github.com/uigraph-oss/uigraph-cli. These are legitimate vendor-provided resources (uigraph-oss) necessary for the skill's primary functionality.
Audit Metadata