call-for-speakers-digest

Pass

Audited by Gen Agent Trust Hub on Aug 30, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [Indirect Prompt Injection Surface]: The skill is designed to ingest and summarize content from multiple external conference aggregators and general web search results.
  • Ingestion points: Data is fetched from sources including confs.tech, papercall.io, adatosystems.com, cfpradar.dev, sessionize.com, communitydays.org, and run.events (as detailed in SKILL.md).
  • Boundary markers: The skill does not define explicit delimiters or instructions to ignore embedded commands within the ingested text.
  • Capability Inventory: The agent has the capability to send emails via Gmail/Outlook and write HTML files.
  • Sanitization: The skill lacks explicit sanitization or filtering logic for the content fetched from third-party sites.
  • Security Consideration: Ingesting data from third-party sites creates a surface for indirect prompt injection, where an attacker could place malicious instructions on a conference page to influence the agent's behavior. This represents a potential risk factor that warrants review.
  • [Dynamic Code Execution]: The skill uses the platform's Code Interpreter to generate the HTML email digest.
  • Evidence: SKILL.md (Step 5) specifies that "Building the HTML digest in Copilot Studio specifically requires Code Interpreter enabled on the agent."
  • Security Consideration: While Code Interpreter is a standard tool for document formatting and data processing, it involves generating and running code based on collected data. This is a common pattern for digest skills but relies on the underlying platform's security sandboxing for safety.
  • [User Data Management]: The skill collects and retains a persistent speaker profile to tailor its searches.
  • Evidence: SKILL.md (Step 1) instructs the agent to "Ask for, and remember, this profile so it doesn't need to be re-collected every week," including topics of interest and travel willingness.
  • Security Consideration: The storage of user preferences is necessary for the recurring nature of the skill. Users should be aware that this information is maintained within the agent's memory to facilitate the automated search functionality.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 30, 2026, 11:55 PM
Security Audit — agent-trust-hub — call-for-speakers-digest