Back to skill

Security audit

Tandoor Recipe CLI

Security checks for vulnerabilities and agentic risk

Overview

The skill is openly designed to manage a Tandoor Recipe Manager instance, but its setup path can run an unpinned npm CLI while granting access that may modify or delete Tandoor data.

Review and pin the exact tandoor-cli version before installing or running it, preferably matching the skill version. Use a short-lived read-only token unless you truly need writes, avoid admin or space-owner tokens for normal recipe tasks, and verify that your agent asks before any write, delete, bulk, household, or invite operation.

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
references/SETUP.md:17
Finding

Unpinned Third-Party CLI Installation and Remote Execution

Content
View full analysis

Vulnerability Details

File Location: references/SETUP.md, lines 17–28
Vulnerability Type: Unpinned third-party dependency and mutable remote package execution
Risk Level: Medium

Vulnerable Code

markdown
### Option 1: Install globally via npm

```bash
npm install -g tandoor-cli

After the global install, the tandoor command becomes available system-wide.

Option 2: Run without installing

bash
npx tandoor-cli <command>

No global install is required. npx downloads and runs the CLI on demand.

text

Additional unpinned fallback commands appear at `references/SETUP.md:231-232`:

```markdown
- Try `npx tandoor-cli` instead
- Reinstall: `npm install -g tandoor-cli`

The documentation also makes an inaccurate version-pinning assertion at SECURITY.md:34:

markdown
- **Version Pinning:** The skill explicitly uses `npx tandoor-cli@0.3.1` to pin to a known version

No corresponding pinned command appears in the reviewed setup instructions. In addition, SKILL.md:5 declares version 1.5.0, making the 0.3.1 security assertion inconsistent with the documented skill version.

Technical Analysis

Both npm install -g tandoor-cli and npx tandoor-cli <command> resolve a mutable package version from the npm registry. The npx command may download and execute package code on demand without constraining it to a previously reviewed release.

The executable implementation of tandoor-cli is not included in the audited artifact. Consequently, the runtime behavior of the downloaded package cannot be verified by this audit. Verification links and recommendations to pin a version do not technically enforce package identity, version, or integrity.

This creates a supply-chain boundary in which the effective executable can change after the skill itself has been reviewed. A compromised maintainer account, malicious package release, registry compromise, or unexpected incompatible update could cause arbitrary pa ...[truncated 1778 chars]

Remediation
View remediation

Remediation Suggestions

  1. Replace every unpinned installation and execution command with the exact reviewed version matching the skill metadata, for example:

    bash
    npm install -g tandoor-cli@1.5.0
    npx --yes tandoor-cli@1.5.0 <command>
    
  2. Update the fallback commands at references/SETUP.md:231-232 so they also specify the exact version.

  3. Correct or remove the inaccurate assertion in SECURITY.md:34. The security documentation, setup guide, and SKILL.md metadata must identify the same reviewed version.

  4. Prefer a project-local dependency managed by a committed lockfile rather than a global installation or unconstrained on-demand execution.

  5. Verify package integrity and provenance before execution. Where supported, use npm integrity metadata, package signatures or attestations, and a trusted internal registry or approved package mirror.

  6. Review the exact CLI release and its transitive dependency tree before granting it Tandoor credentials, especially administrative or space-owner credentials.

  7. Run the CLI under a restricted operating-system account or isolated environment with access only to the minimum required files and environment variables.

  8. Supply short-lived, least-privileged Tandoor tokens only for the duration of the required operation. Do not expose administrative credentials during ordinary recipe or shopping-list tasks.

Vulnerability Patterns
  • 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
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Rogue AgentSelf-Modification, Session Persistence
Findings (17)

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · SECURITY_REVIEW_RESPONSE.md (reported line 135)May include surrounding context.

md
## 4. Identity and Privilege Abuse - Long-Lived Tokens (MEDIUM SEVERITY)

### Original Concern
> The setup guide demonstrates creating a long-lived read/write OAuth2 access token, which would let the CLI perform both read and mutation operations for up to a year if not revoked.

### Resolution

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/SETUP.md (reported line 93)May include surrounding context.

md
### API token (preferred)

Use an OAuth2 access token with Bearer authentication. 

**Important:** The regular DRF token shown in Tandoor's Settings → API does **not** work for this CLI. You need an OAuth2 access token.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/SETUP.md (reported line 95)May include surrounding context.

md
### API token (preferred)

Use an OAuth2 access token with Bearer authentication. 

**Important:** The regular DRF token shown in Tandoor's Settings → API does **not** work for this CLI. You need an OAuth2 access token.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/SETUP.md (reported line 133)May include surrounding context.

expires=timezone.now() + timedelta(days=7), # 7 days scope='read', # Read-only ) print('ACCESS TOKEN:', token.token) "

text

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/SETUP.md (reported line 167)May include surrounding context.

expires=timezone.now() + timedelta(days=7), # 7 days scope='read', # Read-only ) print('ACCESS TOKEN:', token.token) "

text

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SECURITY.md (reported line 7)May include surrounding context.

md
## Overview

The tandoor-cli skill provides programmatic access to your Tandoor Recipe Manager instance. Like any tool with write access to your data, it requires careful configuration and usage to maintain security.

## Addressed Security Concerns

Unsafe Defaults

Medium
Category
Tool Misuse
Confidence
70% confidence
Finding

Tool defaults are unsafe or overly permissive (e.g. disabled TLS verification, no authentication, world-writable permissions). Unsafe defaults widen the attack surface.

Content

Scanner excerpt · SECURITY.md (reported line 152)May include surrounding context.

md
### ❌ Insecure Methods

- **Plain text in code:** Never commit tokens to version control
- **Shared config files:** Avoid world-readable files
- **Chat logs:** Don't paste tokens in Slack, Discord, etc.
- **Browser history:** Be careful with tokens in URLs

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SECURITY_REVIEW_RESPONSE.md (reported line 29)May include surrounding context.

md
1. **Command Classification System**
   - All commands now categorized by risk level:
     - Read operations (safe, no approval)
     - Write operations (require user confirmation)
     - Destructive operations (require explicit confirmation)
     - Administrative operations (require privileged token + explicit confirmation)

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The document explicitly acknowledges that use of npm/npx may fetch and execute external package code on demand, and the response only adds documentation and recommendations rather than a technical control. In the context of an agent skill that may invoke a CLI automatically, an unpinned package introduces a real supply-chain risk because the executed code can change over time or be substituted upstream.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 31)May include surrounding context.

md
**⚠️ IMPORTANT: This skill provides mutation authority over your Tandoor instance.**

- **Read operations** (list, search, get, random) are safe and require no confirmation
- **Write operations** (add, update, import) modify data but are typically reversible
- **Destructive operations** (delete, clear, household management) require explicit user approval before execution
- **Bulk operations** (shopping check --all, clear) affect multiple items and require confirmation

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SECURITY.md (reported line 221)May include surrounding context.

md
**Before executing any command, the agent MUST:**

1. **For read operations:** Proceed without confirmation
2. **For write operations:** Describe the action and wait for user approval
3. **For destructive operations:** 
   - Clearly explain what will be deleted/modified

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SECURITY_REVIEW_RESPONSE.md (reported line 213)May include surrounding context.

md
**Before executing any command, the agent MUST:**

1. **For read operations:** Proceed without confirmation
2. **For write operations:** Describe the action and wait for user approval
3. **For destructive operations:** 
   - Clearly explain what will be deleted/modified

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 179)May include surrounding context.

md
**Before executing any command, the agent MUST:**

1. **For read operations:** Proceed without confirmation
2. **For write operations:** Describe the action and wait for user approval
3. **For destructive operations:** 
   - Clearly explain what will be deleted/modified

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 180)May include surrounding context.

md
**Before executing any command, the agent MUST:**

1. **For read operations:** Proceed without confirmation
2. **For write operations:** Describe the action and wait for user approval
3. **For destructive operations:** 
   - Clearly explain what will be deleted/modified
   - Warn that the action cannot be easily reversed

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.