Back to skill

Security audit

plandeck

Security checks for vulnerabilities and agentic risk

Overview

Plandeck is a coherent planning board, but it needs review because its optional board server can expose plan details without authentication if bound beyond localhost.

Install only if you are comfortable with a local planning tool that writes tracking files beside your plan. Keep the board bound to localhost unless you have added authentication or network controls, and avoid placing secrets in cards, receipts, or note files because the board renders those details.

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

Warning
Location
scripts/board.mjs:52
Finding

Unauthenticated Information Disclosure Through Non-Loopback Board Binding

Content
View full analysis
{ handleRequest(hub, req, res).catch((error) => { if (res.headersSent) { res.end(); return; } sendText(res, error.statusCode || 400, error.message || "Request failed", { closeConnection: error.closeConnection === true, socket: req.socket, }); }); }); server.listen(candidate, host, () => { server.off("error", fail); ok(); }); ``` Board and journal contents are returned through unauthenticated GET endpoints: ```js function handleBoardRequest(record, req, res, pathname, searchParams) { const prefix = record.boardPath.slice(0, -1); if (pathname === prefix) { redirect(res, record.boardPath); return; } if (req.method !== "GET") { methodNotAllowed(res, "GET"); return; } const rest = pathname === record.boardPath ? "/" : pathname.slice(prefix.length); if (rest === "/api/board") { sendJson(res, record.lastPayload); return; } if (rest === "/api/journal") { sendJson(res, { entries: readJournal(record.root, { since: searchParams.get("since"), limit: Number(searchParams.get("limit")) || 2 ...[truncated 4421 chars]
Remediation
View remediation

T08 · Insecure Dependencies

Note
Location
README.md:8
Finding

Unpinned Installer and Repository Sources in Recommended Installation Command

Content
View full analysis
Remediation
View remediation
add ... ``` 2. Pin the GitHub source to an immutable commit hash or signed release tag supported by the installer. 3. Publish checksums or signatures for released Skill artifacts and document how users can verify them. 4. Prefer a package-manager workflow with a lockfile or integrity metadata where practical. 5. Document that `npx` may download and execute code, and recommend inspecting the resolved package and Skill revision before installation in sensitive environments. 6. Maintain a release page that maps each published package version to its exact Git commit and expected artifact digest. ]]>
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Memory PoisoningPersistent Context Injection, Context Window Stuffing, Memory Manipulation
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Findings (16)

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 · README.md (reported line 70)May include surrounding context.

md
Plan slugs stay readable. If two different roots both use `slug: launch`, their paths become `/launch/` and `/launch-2/`. Registering the same root twice refreshes the existing board. When a plan disappears, the hub closes its watcher and SSE clients, then removes it from the index. Every occupied port in the Windows-safe ladder is checked for a compatible hub before Plandeck moves on.

Security posture: the hub binds to `127.0.0.1` by default. `POST /api/boards` and `DELETE /api/boards` require `Content-Type: application/json` and an exact Host header for `127.0.0.1`, `localhost`, or `plandeck.localhost` on the active port. Other hosts are rejected. The temporary breadcrumb is only a discovery hint and is never trusted without the hub handshake.

## The board model

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 104)May include surrounding context.

md
Plan slugs stay readable. If two different roots both use `slug: launch`, their paths become `/launch/` and `/launch-2/`. Registering the same root twice refreshes the existing board. When a plan disappears, the hub closes its watcher and SSE clients, then removes it from the index. Every occupied port in the Windows-safe ladder is checked for a compatible hub before Plandeck moves on.

Security posture: the hub binds to `127.0.0.1` by default. `POST /api/boards` and `DELETE /api/boards` require `Content-Type: application/json` and an exact Host header for `127.0.0.1`, `localhost`, or `plandeck.localhost` on the active port. Other hosts are rejected. The temporary breadcrumb is only a discovery hint and is never trusted without the hub handshake.

## The board model

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 94)May include surrounding context.

md
From the repo root, run each command with `node scripts/cli.mjs <cmd>`, or as `plandeck <cmd>` after `npm link` (a published `npx plandeck` release is planned).

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 · scripts/board.mjs (reported line 287)May include surrounding context.

js
if (req.method === "DELETE") {
      requireTrustedMutation(hub, req);
      const boardPath = normalizeBoardPath(requestUrl.searchParams.get("path"));
      if (!boardPath) throw requestError(400, "DELETE /api/boards needs a board path.");
      if (!removeBoard(hub, boardPath)) {
        sendJson(res, { removed: false, boardPath }, 404);
        return;

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 · scripts/board.mjs (reported line 732)May include surrounding context.

js
const listed = boards.length
    ? boards.map((record) => `- ${record.boardPath}  ${record.lastPayload?.plan?.title || "Plandeck"}`).join("\n")
    : "- (none)";
  sendText(res, 404, `Unknown Plandeck hub path: ${pathname}\n\nRegistered boards:\n${listed}\n\nDo not kill the hub to restart one board. Unregister that board with DELETE /api/boards?path=/<slug>/.`);
}

function sortedBoards(hub) {

Memory Manipulation

High
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.

Content

Scanner excerpt · scripts/lib/journal.mjs (reported line 78)May include surrounding context.

js
.filter((entry) => entry.cardId)
      .sort(compareEntries);
    if (entries.length) appendEntries(planDir, entries);
    for (const entry of entries) delete state.cards[entry.cardId];
    writeLastStateUnsafe(planDir, state);
  } catch (error) {
    warnOnce(planDir, error);

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.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · README.md (reported line 200)May include surrounding context.

md
priority: P1
    risk: med
    depends_on: [C002]
    verify: ["curl -f localhost:3000/api/signup -d @test/fixtures/user.json"]
    next_action: "Wire the POST /signup handler to the users table, return 201 with the new id."
    tags: [api, auth]
    receipt:

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
84% confidence
Finding

The skill advertises shell/CLI behavior that uses environment variables and exposes a localhost network service, but it declares no explicit tool scope or permission boundaries in the skill manifest. In an agent ecosystem, missing scope increases the chance the skill is invoked with broader capabilities than intended, making file, env, and local-network interactions less constrained and harder to audit.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The trigger list contains broad phrases like planning, task board, and 'what is next', which are common in ordinary conversations and can cause accidental invocation. Unintended activation is risky here because the skill can lead an agent to run local commands, read/write plan files, and potentially start a localhost server, expanding actions beyond the user's actual intent.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
74% confidence
Finding

Referencing a future or generic npx plandeck install path without a pinned version allows whatever package version resolves at execution time to run. If the package is compromised, typosquatted, or updated maliciously, an agent could fetch and execute unreviewed code with the agent's local privileges.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

This YAML file describes agent interaction with the plan as a general capability, but it does not define when the agent should act, what triggers updates, or any exclusion conditions. In a manifest-like file, that lack of specificity can create ambiguity about when automated edits are appropriate.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
97% confidence
Finding

The HTML embeds absolute local filesystem paths in window.__PLANDECK_STATIC__, including a Windows username and document locations. If this preview file is shared, published, or served to other users, it leaks environment details that can aid user identification, targeted social engineering, or reveal internal directory structure unnecessarily.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The code uses localeCompare(..., "en", ...), which forces English collation behavior. This is a natural-language/locale policy concern because the file does not offer a locale choice or document why English-only ordering is required.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The generated page sets <html lang="en">, which forces an English locale without any user choice or opt-in. Under the policy rule, fixed language selection is a natural-language locale constraint unless it is user-selectable or clearly justified as region-specific.

Content

No source excerpt is available for this finding.