pp-tella

Warn

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
  • [DATA_EXFILTRATION]: The skill documentation describes a --deliver webhook:<url> flag. This feature allows the agent to transmit command output—which may include sensitive video metadata, workspace statistics, or private transcripts—to any user-defined or attacker-provided URL via a POST request.
  • [CREDENTIALS_UNSAFE]: The skill instructs users to manually extract a TELLA_SESSION_COOKIE (containing __Secure-Tella.session and XSRF-TOKEN) from a logged-in browser to access unofficial API features. Handling raw session cookies in environment variables increases the risk of account takeover and credential theft if the environment is compromised.
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of the tella-pp-cli binary. It provides installation commands using npx from @mvanhorn/printing-press-library and go install from a GitHub repository managed by the vendor. While these are from the skill's own vendor, they represent external code dependencies executed on the host.
  • [INDIRECT_PROMPT_INJECTION]: The skill acts as a surface for indirect prompt injection by processing external data such as video transcripts and webhook messages. Malicious content embedded in a video's transcript could potentially influence the agent's behavior when it performs a transcripts search or summarizes video content.
  • Ingestion points: Untrusted data enters the context via transcripts search, webhooks tail, and videos get outputs in SKILL.md.
  • Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands within the ingested data.
  • Capability inventory: The CLI tool can write to the local file system (--deliver file:), send data over the network (--deliver webhook:), and perform API mutations.
  • Sanitization: The tool uses structured JSON output and field filtering (--select), which helps organize data but does not sanitize the text content for potential injection attacks.
  • [COMMAND_EXECUTION]: The skill is primarily designed to execute local shell commands via the tella-pp-cli binary. While this is the intended purpose, the agent has the capability to run a wide range of subcommands that mutate data and interact with the network.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 3, 2026, 11:43 AM