Back to skill

Security audit

Biotech Pitch Deck Narrative

Security checks for vulnerabilities and agentic risk

Overview

This skill is a local biotech pitch-deck writing aid with minor packaging and quality issues, but no evidence of hidden, destructive, or data-exfiltrating behavior.

Before installing, consider removing the two requirements.txt entries or installing in an isolated environment. Do not feed confidential biotech, fundraising, or non-public company information unless you are comfortable with the local agent and workspace handling that data.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T08 · Insecure Dependencies

Warning
Location
requirements.txt:1
Finding
Unnecessary and Unpinned Third-Party Dependencies## Vulnerability Details **File Location**: `requirements.txt`, lines 1–2 **Vulnerability Type**: Unnecessary unpinned dependencies and avoidable supply-chain exposure **Risk Level**: Medium **Complete Code Snippet**: ```text dataclasses enum ``` ### Technical Analysis The project declares `dataclasses` and `enum` as external dependencies without exact version constraints or cryptographic hashes. The project documentation requires Python 3.7 or later, where both `dataclasses` and `enum` are already available from the Python standard library. Consequently, these package-index dependencies are unnecessary. Installing them causes a package manager to resolve mutable third-party distributions whose contents are not controlled by this repository. Because neither versions nor hashes are specified, a later installation may retrieve different artifacts from those reviewed during this audit. This finding does not establish that the currently published packages are malicious. It identifies an avoidable supply-chain trust boundary that could become exploitable if a package release, package-index account, mirror, or dependency-resolution source were compromised. ### Attack Path 1. A user or automated deployment process executes `pip install -r requirements.txt`. 2. The package manager queries its configured package index or mirror for `dataclasses` and `enum`. 3. Because no versions or hashes are pinned, the resolver accepts matching mutable distributions selected at installation time. 4. If an accepted distribution has been compromised or substituted, attacker-controlled package content enters the Python environment. 5. Malicious package behavior could then run during installation or when the package is imported by this or another process in the same environment. ### Impact Assessment Successful exploitation could execute code with the privileges of the account performing package installation or running the affected Python ...[truncated 379 chars]
Remediation
## Remediation Suggestions 1. Remove both dependency declarations because Python 3.7 and later provide the required modules through the standard library. 2. Delete `requirements.txt` if the project has no external runtime dependencies, or leave it explicitly empty with a clarifying comment. 3. If third-party dependencies are added later, pin exact versions and use cryptographic hashes, such as with `pip install --require-hashes`. 4. Generate and review a reproducible lock file through an approved dependency-management workflow. 5. Restrict installations to a trusted package index or internal mirror and incorporate dependency scanning into CI. 6. Install dependencies in an isolated, least-privileged virtual environment or build container.
Vulnerability Patterns
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (3)

Natural-Language Policy Violations

Low
Confidence
95% confidence
Finding
The markdown states a known issue that the interface help text is in Chinese and plans to translate all interface text to English. This implies the skill currently forces a specific language/locale for at least part of the user experience, without offering a user choice or documenting a region-specific reason.

Unpinned Dependencies

Low
Category
Supply Chain
Content
dataclasses
enum
Confidence
60% confidence
Finding
Dependencies lack version pinning, allowing potential malicious package updates. Consider pinning versions.

Unpinned Dependencies

Low
Category
Supply Chain
Content
dataclasses
enum
Confidence
60% confidence
Finding
Dependencies lack version pinning, allowing potential malicious package updates. Consider pinning versions.

Static analysis

No suspicious patterns detected.