Back to skill

Security audit

Firefly III API Client

Security checks for vulnerabilities and agentic risk

Overview

This is a legitimate Firefly III API client, but it exposes broad financial mutation/admin authority and can send bearer tokens over HTTP unless the user avoids the documented example.

Install only if you intend to give the agent broad Firefly III API power. Use HTTPS for FIREFLY_URL, keep the token least-privileged if possible, and require explicit human confirmation before exports, bulk edits, deletes, cron, webhook, or user-administration calls.

Vulnerability Patterns
  • 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
  • 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
Findings (2)

T09 · Insecure Skill Coding Practices

Error
Location
src/api.py:8
Finding

Bearer Token May Be Transmitted over Plaintext HTTP

Content
View full analysis
` header. 4. The client sends the authenticated request over plaintext HTTP. 5. An attacker on the same network path captures the token using traffic inspection or a man-in-the-middle position. 6. The attacker reuses the bearer token against the Firefly III API. 7. Subject to the token holder's API privileges, the attacker reads financial records or invokes sup ...[truncated 669 chars]
Remediation
View remediation

T08 · Insecure Dependencies

Note
Location
SKILL.md:17
Finding

Unpinned Third-Party Dependency Installation

Content
View full analysis
Remediation
View remediation
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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (14)

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 · SKILL.md (reported line 10)May include surrounding context.

See SECURITY.md for information on token safety, environment variables, and network security when using this skill.

Setup

  1. Ensure Firefly III is running and you have an OAuth Personal Access Token.
  2. Set the following environment variables:
    bash
    FIREFLY_URL=http://your-instance-url
    

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 · manifest.json (reported line 14)May include surrounding context.

json
See `SECURITY.md` for information on token safety, environment variables, and network security when using this skill.

## Setup
1. Ensure Firefly III is running and you have an OAuth Personal Access Token.
2. Set the following environment variables:
   ```bash
   FIREFLY_URL=http://your-instance-url

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The webhook submission and trigger endpoints allow the skill to cause asynchronous outbound network requests that may transmit stored transaction or account data to external systems. For an unknown-purpose skill, this is an unjustified exfiltration-capable side effect and expands risk beyond the local financial system.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The API surface includes destructive operations such as data destroy, purge, and permanent deletion of major financial objects. In a skill with no narrowing purpose statement, these capabilities materially increase the chance that an agent, prompt injection, or user misunderstanding causes irreversible loss of financial records.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

The bulk transaction update endpoint can modify many financial records at once, but the description does not provide a strong user-facing safety warning or confirmation expectation. In an agent context, insufficient hazard signaling increases the risk of accidental mass modification through ambiguous prompts or automation errors.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The export endpoints enable download of highly sensitive financial records, yet the spec omits clear privacy warnings about handling, storage, and onward sharing of exported files. In a skill setting, this can normalize silent exfiltration of sensitive data to local storage, logs, or third-party tools.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The spec exposes a cron-execution endpoint that can trigger background jobs for all users when supplied with a CLI token. In an agent skill of unknown purpose, this grants a powerful operational side effect unrelated to normal end-user financial interactions and could lead to duplicate transactions, unexpected automated actions, or service-wide changes if misused.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill includes user-administration endpoints for listing, creating, updating, and deleting users, which is excessive for an unspecified financial assistant integration. Exposing identity-management operations to a general-purpose agent increases the blast radius from data access to full account lifecycle control and privilege misuse.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The webhook submission endpoint may send queued webhook messages containing stored user data to external receivers, but the description does not clearly warn about that external transmission. In an agent-mediated environment, missing disclosure materially increases the chance of unintentional data release.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The transaction-triggered webhook endpoint can transmit transaction data to an external webhook receiver, but the spec does not prominently warn users about this outbound disclosure. That omission is dangerous in a financial skill because an agent may trigger exfiltration without the user appreciating that data is leaving the platform.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
98% confidence
Finding

The summary for this autocomplete endpoint explicitly states 'English only', which imposes a language constraint without user opt-in or documented justification. That conflicts with the language/locale policy criterion for natural-language policy violations.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
98% confidence
Finding

The /v1/attachments/{id}/download documentation states it downloads the binary content of a transaction, but the path and response clearly concern an attachment. This is an active documentation contradiction that could mislead consumers about what object is being accessed.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
97% confidence
Finding

The /v1/rules/{id}/trigger endpoint is a single-rule trigger, yet its description says 'Fire the rule group on your transactions' and mentions rules in the group. That wording contradicts the path and operationId, which target one rule rather than a rule group.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
85% confidence
Finding

This manifest describes the skill's functionality but does not specify any explicit trigger phrases, invocation constraints, or exclusion conditions. For manifest files, missing specificity on when the skill should activate can lead to overly broad or unintended invocation by a host system.

Content

No source excerpt is available for this finding.