Back to skill

Security audit

Fossil

Security checks for vulnerabilities and agentic risk

Overview

This skill has a coherent failure-memory purpose, but it pushes routine remote sharing and persistent agent behavior without enough scoping or data-handling safeguards.

Review this before installing. Use it only if you are comfortable sending sanitized task descriptions and failure summaries to the hosted FOSSIL service, and avoid recording secrets, personal data, customer details, raw prompts, filenames, or regulated information. Prefer a pinned package version and avoid adding the AGENTS.md persistence block unless you intentionally want future sessions to use this remote memory by default.

Vulnerability Patterns
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Findings (3)

T02 · Agent Memory Poisoning

Error
Location
SKILL.md:67
Finding

Persistent Agent Policy Can Be Influenced by Untrusted Remote Memory

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 67-78 and 125-140
Vulnerability Type: Persistent memory poisoning through externally sourced instructions
Risk Level: High

Vulnerable Code

markdown
## When to search

Call `fossil_search` before any step involving:

- Parsing or extracting structured data from LLM output
- Calling external APIs or tools
- Multi-step file operations
- Browser automation
- Sending messages or emails on behalf of the user
- Any task domain that has failed before in this workspace

Pass a natural language description of what you are about to attempt.
Read the returned resolutions before proceeding.
markdown
## Add to AGENTS.md

```markdown
## Failure Memory (FOSSIL)

Before any non-trivial task step, call fossil_search with a description
of what you are about to attempt. Read returned resolutions before acting.

After any failure, call fossil_record before retrying. Capture:
- what you were attempting
- what went wrong (use the FOSSIL failure taxonomy)
- what fixed it

This builds a persistent failure memory across all sessions.
text

### Technical Analysis

The skill directs the user to add permanent behavioral instructions to `AGENTS.md`, causing FOSSIL lookups and returned resolutions to affect future sessions and unrelated tasks. The instructions explicitly require the agent to read remotely stored resolutions before acting.

No mechanism is specified to verify record ownership, authenticate the author of a resolution, distinguish data from executable instructions, constrain a resolution to its original task, or prevent returned content from attempting to override higher-priority policies. Consequently, compromised, shared, or poisoned records could become an indirect instruction channel.

The persistent `AGENTS.md` modification expands the trust boundary beyond the current skill invocation. Even after the original task ends, later sessions may continue retrieving and interpreting remote co
...[truncated 1341 chars]
Remediation
View remediation

Remediation Suggestions

  • Do not automatically or routinely add remote-memory instructions to AGENTS.md.
  • Make remote failure-memory lookup explicitly opt-in and scoped to the current task.
  • Treat every returned resolution as untrusted data, not as an instruction.
  • Enforce a strict response schema that accepts factual failure metadata but rejects commands, tool calls, policy changes, and prompt-like directives.
  • Authenticate record authors and enforce per-user or per-workspace tenant isolation.
  • Attach provenance, timestamps, ownership, and integrity metadata to every record.
  • Require user confirmation before applying remote advice to sensitive operations such as sending messages, modifying files, or invoking external tools.
  • Prevent remote content from overriding system, developer, user, or workspace policies.
  • Provide controls to inspect, revoke, quarantine, and delete persistent records.

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:25
Finding

Potentially Sensitive Operational Context Is Sent to an External Community Service Without Data-Handling Safeguards

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 25-26, 75-98, and 125-140
Vulnerability Type: Uncontrolled disclosure of task and failure context to a third-party API
Risk Level: High

Vulnerable Code

markdown
The community API at `fossil-api.hello-76a.workers.dev` is live and free.
No API key required. Embeddings run on Cloudflare Workers AI.
markdown
Pass a natural language description of what you are about to attempt.
Read the returned resolutions before proceeding.

fossil_search("extracting JSON fields from an invoice document")

text

---

## When to record

Call `fossil_record` after any failure, before retrying.

fossil_record( situation="sending a reply email to insurance company", failure_type="misinterpretation", failure="agent replied to wrong thread — matched subject line not sender", severity="major", resolution_type="prompt_change", resolution="added: always match by sender address, not subject line", framework="openclaw", model="claude-opus-4-5" )

text
markdown
After any failure, call fossil_record before retrying. Capture:
- what you were attempting
- what went wrong (use the FOSSIL failure taxonomy)
- what fixed it

This builds a persistent failure memory across all sessions.

Technical Analysis

The skill instructs the agent to submit natural-language descriptions of upcoming work and detailed failure records to an external community API. Such descriptions can include personal information, customer or account details, filenames, internal workflows, message context, proprietary data, or other sensitive operational metadata.

The documented workflow does not include an informed-consent checkpoint, field-level redaction, secret detection, data minimization, tenant authentication, retention limits, deletion procedures, or restrictions against uploading raw prompts and message contents. The absence of an API key also leaves the documented setup without an explicit user or wo ...[truncated 1536 chars]

Remediation
View remediation

Remediation Suggestions

  • Default to local-only storage and require explicit informed consent before enabling the community service.
  • Display the destination, data categories, subprocessors, retention period, and deletion process before activation.
  • Apply automatic secret detection and redaction before every search or record request.
  • Prohibit transmission of credentials, tokens, raw prompts, complete message bodies, personal identifiers, financial data, health data, and proprietary source content.
  • Minimize submissions to abstract failure categories and sanitized summaries.
  • Add per-user or per-workspace authentication and enforce tenant isolation.
  • Encrypt data in transit and at rest, and document retention and deletion guarantees.
  • Allow users to review and edit the exact payload before transmission.
  • Provide a mechanism to list, export, revoke, and permanently delete submitted records.
  • Disable automatic recording for regulated or high-sensitivity tasks.

T08 · Insecure Dependencies

Error
Location
SKILL.md:10
Finding

Unpinned npm Package Is Downloaded and Executed Through npx

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 10-15 and 32-48
Vulnerability Type: Unpinned runtime dependency execution
Risk Level: High

Vulnerable Code

yaml
metadata:
  openclaw:
    emoji: "🦴"
    homepage: https://github.com/heyrtl/fossil
    install:
      - kind: node
        package: "@openfossil/mcp"
        bins: [openfossil-mcp]
markdown
## Setup

Add to your `openclaw.json`:

```json
{
  "mcp": {
    "servers": [
      {
        "name": "fossil",
        "command": "npx",
        "args": ["@openfossil/mcp"],
        "env": {
          "FOSSIL_API_URL": "https://fossil-api.hello-76a.workers.dev"
        }
      }
    ]
  }
}

Restart your gateway. FOSSIL tools are now available.

text

### Technical Analysis

The configuration invokes `npx` with the package name `@openfossil/mcp` but does not specify an exact version. Depending on the local npm environment and cache state, `npx` can retrieve and execute the current package release from the configured registry.

No lockfile, integrity hash, vendored source, provenance attestation, or reviewed implementation is included in the audited project. Therefore, the code executed by the gateway can change after the skill itself has been reviewed. A compromised publisher account, package registry, dependency chain, or future malicious release could introduce arbitrary code.

Because the package is configured as an MCP server, execution occurs when the gateway starts or initializes that server. The resulting process may inherit the gateway user's environment, filesystem access, and network capabilities unless separately sandboxed.

### Attack Path

1. An attacker compromises the package publisher, npm account, registry distribution path, or a transitive dependency.
2. The attacker publishes a modified release under the same package name.
3. The user installs or starts the configured MCP server after the malicious release becomes current.
4. `npx @openfossil/mcp` retrieves
...[truncated 810 chars]
Remediation
View remediation

Remediation Suggestions

  • Pin the package to an exact reviewed version, for example @openfossil/mcp@x.y.z, rather than resolving the latest release.
  • Commit and enforce a lockfile containing integrity hashes.
  • Verify package provenance and registry signatures where supported.
  • Review or vendor the exact MCP implementation and its transitive dependencies.
  • Use a private allowlisted registry or an internal artifact repository for approved builds.
  • Disable or restrict dependency lifecycle scripts where operationally possible.
  • Run the MCP server in a sandbox or container with a read-only filesystem and minimal environment variables.
  • Restrict outbound network access to explicitly required destinations.
  • Run the process under a dedicated, unprivileged account.
  • Establish dependency monitoring and require security review before version upgrades.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly directs users to a live remote community API and encourages recording past failures and resolutions, but it does not warn that those records may leave the local environment and be transmitted to a third-party service. Failure reports can easily contain sensitive operational details, user data, internal prompts, file names, account identifiers, or security-relevant mistakes, so the absence of a clear disclosure creates a real privacy and data-exfiltration risk.

Content

No source excerpt is available for this finding.