make-app-actions

Warn

Audited by Gen Agent Trust Hub on Aug 28, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill directs the agent to dynamically discover and ingest instructions from a package.ai.json file located within the node_modules/@qfei-design/make-app-actions directory. It instructs the agent to parse a readOrder property to determine which files to read to understand the package API. This pattern allows external content from a dependency to influence agent behavior at runtime, which could be exploited if the dependency is compromised or typosquatted.
  • Evidence: SKILL.md (Workflow Step 2) and references/package-integration.md (Pre-flight section).
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of data that can be influenced by external actors, such as record fields, search filters, and group expressions. These values are ingested and interpolated into service requests and UI displays without explicit instructions for using boundary markers to prevent the agent from interpreting embedded instructions.
  • Ingestion points: Untrusted data enters via loadedRecords in references/selection-and-operation-snapshot.md and filter/groupFilter expressions in SKILL.md.
  • Boundary markers: The instructions do not define delimiters or warnings to ignore embedded instructions in the processed data.
  • Capability inventory: The skill performs record mutations via POST /api/make/app/entities/:objectKey/records/bulk as described in references/service-contract.md.
  • Sanitization: While the skill mentions structural validation of filters in references/service-contract.md, it lacks prompts for sanitizing natural language content to prevent injection attacks.
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of specific Node.js packages: @qfei-design/make-app-actions and @qfei-design/canvas-table. These packages are vendor-owned resources associated with the skill author 'qfeius' and are necessary for the skill's primary function of integrating record actions into a CanvasTable environment.
  • Evidence: references/package-integration.md (Pre-flight section).
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 28, 2026, 07:07 AM
Security Audit — agent-trust-hub — make-app-actions