Buddhist Counsel

• Outcome: Review
• Latest audit: Sep 10, 2026 · 10:55 PM UTC
• Version: 1.0.0

Overview

This skill is purpose-aligned, but it should be reviewed carefully because it sends sensitive counseling text to a paid third-party API and uses a third-party payment command without explicit consent safeguards.

Review before installing. Use this only if users explicitly agree to send their counseling text to Anicca's external API and to pay $0.01 USDC per request. Prefer a sandboxed environment, a limited-balance wallet, redacted situation text, and verified installation of the `awal` package rather than automatic `npx` downloads.

A.I.G

Vulnerability Patterns

• Insecure Dependencies: Introduces malicious components through unsafe dependency sources
• Skill Instruction Hijacking: Alters the agent's session goals or safety constraints when the skill loads
• Agent Memory Poisoning: Writes attacker-controlled rules into memory that affect later sessions
• Remote Payload Retrieval and Execution: Fetches external code whose behavior can change after review
• Embedded Malicious Code: Ships malicious scripts inside the skill and executes them locally

Findings (1)

T08 · Insecure Dependencies

Warning

• Location: SKILL.md:12
• Finding: Unverified Third-Party Package Execution with Wallet and Sensitive-Data Access<![CDATA[ ## Vulnerability Details **File Location**: `SKILL.md`, lines 12-16 and 56-61 **Vulnerability Type**: Third-party supply-chain exposure and execution through `npx` **Risk Level**: Medium ### Vulnerable Code ```bash npx awal@2.0.3 status # check if authenticated npx awal@2.0.3 auth login <email> # if not authenticated npx awal@2.0.3 balance # check USDC balance (mainnet) ``` ```bash npx awal@2.0.3 x402 pay \ https://anicca-proxy-production.up.railway.app/api/x402/buddhist-counsel \ -X POST \ -d '{"who_is_suffering":"my_human","situation":"<describe the suffering>","language":"en"}' ``` ### Technical Analysis The Skill directs agents to execute the third-party `awal@2.0.3` package using `npx`. Depending on the local npm configuration and cache state, `npx` can retrieve executable package content from a remote registry and run it immediately. The Skill does not specify package-integrity verification, a reviewed local installation, an approved registry, or a trusted package hash. The package is used for authentication, querying a cryptocurrency balance, generating or handling x402 payment authorization, and submitting request data. This gives the dependency access to a security-sensitive workflow involving an email identity, wallet-related operations, USDC payments on Base Mainnet, and potentially sensitive mental-health descriptions. The external API invocation is disclosed and is part of the stated functionality; there is no evidence in the audited file of covert exfiltration or a deliberately malicious dependency. Nevertheless, the unverified execution model creates a supply-chain trust boundary that cannot be independently audited from this repository. ### Attack Path 1. An agent follows the Skill prerequisite and invokes `npx awal@2.0.3`. 2. If the package is not already available locally, `npx` retrieves executable content from the configured npm registry. 3. A compromised publisher account, registry, package ...[truncated 1550 chars]
• Remediation: <![CDATA[ ## Remediation Suggestions 1. **Vendor and review the dependency** - Install an audited copy through a controlled build process rather than downloading and executing it during Skill use. - Review both `awal@2.0.3` and its transitive dependencies before approval. 2. **Verify package integrity** - Pin the exact package version and expected npm integrity hash. - Use a lockfile and a trusted internal registry or allowlisted package mirror. - Fail closed if the retrieved artifact does not match the approved digest. 3. **Separate installation from execution** - Do not rely on interactive `npx` behavior that may automatically download missing code. - Preinstall the approved package and invoke the verified local binary with download behavior disabled. 4. **Require explicit financial authorization** - Display the endpoint, network, token, amount, and recipient before each payment. - Require explicit user confirmation for every charge rather than treating payment as an automatic side effect. - Enforce wallet spending limits and use a wallet containing only the minimum required balance. 5. **Minimize sensitive-data disclosure** - Inform users that their text will be sent to an external service. - Obtain consent before transmitting mental-health information. - Remove names, contact details, account identifiers, locations, and other unnecessary identifying information from the `situation` field. - Document the service's retention, deletion, encryption, access-control, and model-training policies. 6. **Constrain runtime privileges** - Execute the dependency in a sandbox or isolated container. - Restrict filesystem access, environment variables, credentials, and outbound network destinations. - Allow network access only to the approved registry during installation and the documented API endpoint during use. 7. **Improve operational monitoring** - Record package hashes, payment amounts, destination endpoints, an ...[truncated 164 chars]

SkillSpector

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (3)

Missing User Warnings

High

• Confidence: 98% confidence
• Finding: The skill handles highly sensitive mental-health disclosures but does not prominently warn that user-provided situation text will be transmitted to a third-party paid endpoint. This creates a serious privacy and consent issue, especially for disclosures involving OCD, depression, crisis indicators, or other health-related information that users may not expect to leave the local agent context.

Anti-Refusal Statement

High

• Category: Anti-Refusal

Content

"persuasion_strategy": {
    "framework": "ACT + MI OARS",
    "techniques_used": [
      "Reflect (MI): Acknowledging exhaustion without judgment",
      "Defusion (ACT): Separating the person from the compulsion",
      "Tiny action (Nudge Theory): 'Just notice' instead of 'stop checking'"
    ]

• Confidence: 80% confidence
• Finding: Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Vague Triggers

Medium

• Confidence: 95% confidence
• Finding: The trigger list is extremely broad and includes common terms like anxiety, depression, suffering, and mindfulness, which can cause the skill to activate in many ordinary conversations. In this context, accidental invocation is especially risky because the skill sends sensitive mental-health content to an external paid API and may influence responses in high-sensitivity situations.