Back to skill

Security audit

Dumb Panel Task CLI

Security checks for vulnerabilities and agentic risk

Overview

This task manager needs review because it borrows panel users’ identities and allows authenticated actions beyond task management.

Review before installing on a live panel. Prefer authentication tied to an explicitly authorized account, restrict raw requests to task endpoints, and reject expired sessions. Jobs may run with powerful service privileges, and --disabled does not create them atomically disabled. All bundled files were inspected, but backend authorization and session-expiration enforcement remain unverified.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • 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
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (12)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The documented raw API capability is broader than the task-management purpose and exposes operations outside that purpose using the locally generated authentication token. Server-side authorization still applies, but a privileged session could make unintended non-task operations available; the documented run and stop commands alone are not evidence of abuse.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding

Updating task metadata and toggling its enabled state match the description. The separate run and stop commands expose execution-control capabilities beyond the declared creation, editing, deletion, and enable/disable operations. The absence of create and delete examples is not itself a mismatch.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding

Listing and retrieving tasks and task groups reasonably support the declared management workflow. However, the explicit run and stop commands expose execution-control capabilities not covered by creating, editing, deleting, or toggling cron tasks. The supplied chunk contains usage examples rather than the helper implementation, so this assessment is limited to the capabilities those commands indicate.

Content

No source excerpt is available for this finding.

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · SKILL.md (reported line 13)May include surrounding context.

md
tasks remain: the panel's own Web API (preferred, takes effect immediately), or writing
SQLite directly (requires an Administrator service restart).

Always prefer the API. On `POST`, `PUT` and `DELETE /api/tasks` the server updates the
in-memory scheduler alongside the database, so no elevation and no restart are needed.

## Which route to take

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 61)May include surrounding context.

md
`references/api_reference.md`.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 121)May include surrounding context.

md
- `references/api_reference.md` — endpoint table, task fields, JWT claims, read-only

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · references/api_reference.md (reported line 171)May include surrounding context.

md
| `PUT /tasks/:id` changing name and cron | 200, `next_run_at` recalculated | full-body merge works |
| `PUT /tasks/:id/disable` | 200, `status=0, enabled=false` | correct way to disable |
| `PUT /tasks/:id/enable` | 200, `status=1, enabled=true` | reversible |
| `DELETE /tasks/:id` | 200 | removes the record only; script files are untouched |
| `ddp.exe task run <id>` | exit code 0 | local blacklist alerts for `wmic` / `reg` / `sc` do not affect execution |

When testing against a live panel, use a cron expression that will not fire soon

Credential Access

High
Category
Privilege Escalation
Confidence
91% confidence
Finding

The implementation reads the server-wide signing secret and selects an enabled user's session without binding that identity to the caller, then mints an access token as that user. Although openly documented and not evidence of exfiltration, this authentication shortcut exposes a high-value credential and may turn local file access into privileged panel access, including task execution.

Content

Scanner excerpt · scripts/ddp_task_api.py (reported line 11)May include surrounding context.

python
both halves: it writes the row and registers the job with the live scheduler, so changes
take effect immediately.

Authentication: the panel signs access tokens with Dumb-Panel/.jwt_secret. This script
reads that key plus an unexpired row from user_sessions and signs a short-lived HS256
token with matching claims (token_type must be "access"). The key, jti and token are
never printed or written anywhere.

The raw command extends task CRUD into unrestricted panel API access.

High
Category
Excessive Agency
Confidence
99% confidence
Finding

The raw command forwards caller-controlled methods, paths, and bodies with the generated bearer token without enforcing a task-endpoint boundary. This permits operations beyond task CRUD within the selected session's server-side privileges, potentially including destructive administrative actions if that session is privileged.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env, file_read, network.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

Token generation falls back to an expired session despite documentation requiring an unexpired session.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

When no unexpired session exists, the fallback selects an expired session and signs a token with a fresh one-hour expiration. This violates the documented session-validity requirement and could revive expired access if the backend does not independently enforce session expiration; otherwise, it produces an authentication failure.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% 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 · scripts/ddp_task_api.py (reported line 243)May include surrounding context.

python
p = sub.add_parser("delete", help="delete a task record (not its script file)")
    p.add_argument("id", type=int)
    p.add_argument("--yes", action="store_true", help="skip confirmation")

    for action in ("run", "enable", "disable", "stop"):
        p = sub.add_parser(action, help=f"{action} a task")