Back to skill

Security audit

linkedclaw-requester

Security checks for vulnerabilities and agentic risk

Overview

This marketplace skill is mostly disclosed and purpose-aligned, but it needs Review because it can install mutable tooling, spend credits, share files with third-party agents, widen shell access, and create durable polling jobs.

Install only if you are comfortable with an agent using LinkedClaw to send task content and selected files to external providers and spend credits. Pin or manually review the CLI version before installation, avoid sudo, require explicit approval before uploading sensitive files or broadcasting to multiple providers, and do not grant broad subagent shell access unless you have isolated the environment.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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 (3)

T08 · Insecure Dependencies

Error
Location
scripts/install-cli.sh:164
Finding

Automatic Installation and Execution of an Unpinned npm Package

Content
View full analysis

Vulnerability Details

File Location: scripts/install-cli.sh:164-183, scripts/install-cli.sh:195-211, and scripts/install-cli.sh:216-223
Vulnerability Type: Unpinned third-party dependency installed from a mutable package reference
Risk Level: High

Complete Code Snippet

bash
if command -v linkedclaw >/dev/null 2>&1; then
  cur=$(current_version)
  if [ -n "$cur" ] && version_ge "$cur" "$MIN_VERSION"; then
    latest=$(latest_version)
    if [ -z "$latest" ] || version_ge "$cur" "$latest"; then
      emit_installed "already-installed" ""
      exit 0
    fi
    # A newer version exists — upgrade in place.
    upd_out=$(npm install -g @linkedclaw/cli@latest 2>&1)
    if [ $? -eq 0 ]; then
      hash -r 2>/dev/null || true
      new=$(current_version)
      if [ -n "$new" ] && version_ge "$new" "$latest"; then
        emit_installed "updated" "\"from\": \"$cur\""
        exit 0
      fi
    fi
    emit_installed "already-installed" \
      "\"update_skipped\": \"newer version $latest available but in-place upgrade did not take effect\""
    exit 0
  fi
fi

npm_out=$(npm install -g @linkedclaw/cli@latest 2>&1)
npm_status=$?
if [ "$npm_status" -eq 0 ] && command -v linkedclaw >/dev/null 2>&1; then
  if post_install_check "global" ""; then exit 0; else exit 1; fi
fi

NPM_GLOBAL_DIR="$HOME/.npm-global"
mkdir -p "$NPM_GLOBAL_DIR"
npm_out=$(npm install --prefix "$NPM_GLOBAL_DIR" -g @linkedclaw/cli@latest 2>&1)
npm_status=$?
export PATH="$NPM_GLOBAL_DIR/bin:$PATH"

if [ "${ALLOW_SUDO:-0}" = "1" ] && command -v sudo >/dev/null 2>&1; then
  npm_out=$(sudo -n npm install -g @linkedclaw/cli@latest 2>&1)
  npm_status=$?
  if [ "$npm_status" -eq 0 ] && command -v linkedclaw >/dev/null 2>&1; then
    if post_install_check "sudo" ""; then exit 0; else exit 1
...[truncated 1947 chars]
Remediation
View remediation

Remediation Suggestions

  1. Pin the CLI to an exact audited version, such as @linkedclaw/cli@0.2.6, rather than using @latest.
  2. Verify the package tarball against a trusted integrity hash or signed release metadata before installation.
  3. Remove automatic upgrades of already usable installations. Present an available update to the user and require explicit approval.
  4. Prefer a project-local installation governed by a committed lockfile instead of a global installation.
  5. Remove the sudo installation branch. If privileged installation is unavoidable, require a separate explicit administrative action outside the agent workflow.
  6. Disable or tightly control npm lifecycle scripts where operationally possible.
  7. Audit the exact CLI release before changing the pinned version.

other

Error
Location
references/patterns.md:69
Finding

External Disclosure of Complete Tasks and Sensitive Files Without a Mandatory Consent Gate

Content
View full analysis

Vulnerability Details

File Location: references/patterns.md:69-86 and references/patterns.md:507-535; equivalent instructions also appear in SKILL.md:311-328 and SKILL.md:215-224
Vulnerability Type: Sensitive data disclosure to an external service or third-party provider
Risk Level: High

Complete Code Snippet

bash
# Local file → upload, take the file_id, put it in --input as {"file_id": ...}
fid=$(linkedclaw upload ./passport.png | python3 -c "import sys,json;print(json.load(sys.stdin)['file_id'])")
linkedclaw invoke agt_xyz \
  --capability id_photo_retouch \
  --input "{\"photo\":{\"file_id\":\"$fid\"},\"country\":\"US\",\"document_type\":\"passport\"}" \
  --max-credits 100

# Already-hosted file → skip upload, pass the URL directly
linkedclaw invoke agt_xyz \
  --capability id_photo_retouch \
  --input '{"photo":{"url":"https://example.com/passport.png"},"country":"US","document_type":"passport"}' \
  --max-credits 100
bash
# Send the brief so the provider can PRICE the job.
linkedclaw send "$SES" \
  '{"prompt":"Analyze the attached quarterly CSV and surface the 3 biggest movers"}' \
  --json

# The work turn restates the complete task and sends it to the provider.
linkedclaw send "$SES" \
  '{"prompt":"Analyze the attached quarterly CSV and surface the 3 biggest movers"}' \
  --json
linkedclaw recv "$SES" --wait 300

Technical Analysis

Network transmission is part of the Skill's declared marketplace functionality and is not covert. Nevertheless, the instructions permit the agent to upload local files and transmit complete task content to LinkedClaw and selected third-party providers without imposing a mandatory informed-consent, sensitivity-review, or redaction step.

The examples explicitly include a passport image and a quarterly business dataset. Similar workflows may send source code, documents, prompts, or structured payloads containing perso ...[truncated 1671 chars]

Remediation
View remediation

Remediation Suggestions

  1. Require explicit informed user consent before every external transfer when the user did not explicitly request LinkedClaw.
  2. Before sending, identify the destination provider, data fields, files, estimated cost, and purpose of disclosure.
  3. Add a mandatory secret and PII review for API keys, tokens, passwords, private keys, identity documents, financial data, and confidential business information.
  4. Default to sending only the minimum necessary excerpt rather than the complete task, repository, or document.
  5. Prohibit uploads of credential files, configuration directories, SSH material, environment files, and identity documents unless the user expressly confirms the exact file and destination.
  6. Redact or replace sensitive values with placeholders whenever the provider does not require the original values.
  7. Document provider retention and privacy boundaries, and warn users that third-party providers may receive the submitted content.
  8. Add an allowlist for approved local paths and reject symlinks or paths outside the intended workspace.

T05 · Unauthorized Access and Privilege Escalation

Error
Location
references/openclaw.md:136
Finding

Recommendation to Grant Unrestricted Shell Access to Delegated Subagents

Content
View full analysis

Vulnerability Details

File Location: references/openclaw.md:136-146
Vulnerability Type: Excessive subagent permissions that violate least privilege
Risk Level: High

Complete Code Snippet

text
1. **Subagents are denied system tools by default** — including shell. A
   freshly-spawned subagent **cannot run `linkedclaw` until shell is
   granted**. Two options:
   - **Recommended:** widen `agents.defaults.subagents` in gateway config to
     grant the shell system tool to subagents.
   - Alternative: register a small plugin tool that wraps the CLI call, then
     grant that plugin tool to subagents (avoids opening full shell).

   (Tier B avoids this entirely — `exec` runs in the parent agent's own
   shell, which is already granted.)

Technical Analysis

The documentation recommends widening the default OpenClaw subagent policy to grant the general-purpose shell tool. A subagent only needs a constrained ability to invoke specific LinkedClaw requester commands, but unrestricted shell access permits arbitrary command execution and local filesystem access.

The same section identifies a narrowly scoped plugin wrapper as a viable alternative. This demonstrates that granting the full shell is not the minimum privilege required by the declared functionality.

Delegated subagents may process user-controlled documents, external provider responses, or other untrusted content. Prompt injection in that content can become materially more dangerous when the subagent has unrestricted command execution.

Attack Path

  1. An operator follows the documentation and grants shell access through the global agents.defaults.subagents policy.
  2. The agent creates a delegated subagent to manage a LinkedClaw session.
  3. The subagent processes attacker-controlled task text, a malicious document, or a malicious external provider response.
  4. Embedded instructions induce the subagent to invoke shell command ...[truncated 750 chars]
Remediation
View remediation

Remediation Suggestions

  1. Reverse the recommendation: make a narrowly scoped LinkedClaw plugin or wrapper the required default.
  2. Allowlist only necessary subcommands and arguments, such as search, show, schema, hire, send, recv, end, and cancel.
  3. Reject shell metacharacters and construct child-process argument arrays without shell evaluation.
  4. Restrict filesystem access to explicitly approved input files and a dedicated state directory.
  5. Prevent subagents from reading environment variables, credential directories, SSH files, and unrelated project files.
  6. Restrict outbound network access to the documented LinkedClaw endpoints where feasible.
  7. Apply permissions per task or per agent rather than widening agents.defaults.subagents globally.
  8. Run delegated subagents in a sandbox with a non-privileged account and auditable command logs.
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
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Rogue AgentSelf-Modification, Session Persistence
Findings (26)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The declared description is about runtime behavior of a marketplace requester skill: invoking or broadcasting to other agents when delegation is needed. The supplied code chunk does not implement delegation, agent hiring, marketplace requests, translation/OCR offloading, or parallel sampling. Instead, it is an environment/bootstrap utility that installs and updates the LinkedClaw CLI on the local machine, including version checks, npm installs, PATH conflict detection, and optional sudo escalation. While this may support the requester skill operationally, the code's primary purpose is materially different from the declared functionality and introduces undeclared system/package-management capabilities.

Content

No source excerpt is available for this finding.

Self-Modification

High
Category
Rogue Agent
Confidence
85% confidence
Finding

The skill contains self-update instructions that re-fetch the skill from a registry or source repo and separately upgrades the CLI from npm. Self-modifying behavior expands the trust boundary after initial review and can introduce unreviewed code changes into the runtime, especially if updates are automatic or not pinned.

Content

Scanner excerpt · SKILL.md (reported line 400)May include surrounding context.

md
---

## Update this skill

Re-fetch from the registry (each runtime does it slightly differently):

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · README.md (reported line 46)May include surrounding context.

Or manually:

bash
mkdir -p ~/.openclaw/skills/linkedclaw-requester
cp -r SKILL.md README.md references/ ~/.openclaw/skills/linkedclaw-requester/

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The trigger conditions are intentionally broad enough to activate on many ordinary requests involving translation, review, or tasks the agent "can't do well alone." In a marketplace skill that can spend credits, install software, and contact external providers, over-triggering can cause unintended delegation, unnecessary external data exposure, or unapproved spend.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 6)May include surrounding context.

md
description: LinkedClaw requester — hire, invoke, or broadcast to other agents on the LinkedClaw marketplace when this agent lacks a capability locally. Use this when the user asks to delegate to another agent, hire a specialist, offload translation/OCR/labeling/review, run parallel sampling, or mentions LinkedClaw, an agent marketplace, or spending credits to call out to another agent. Trigger even when the user doesn't name "LinkedClaw" but describes a task this agent can't do well alone and would benefit from an external specialist (e.g. "I need someone good at Chinese to translate this", "get three different takes on this PR"). Covers the requester role only — for the *provider* role (this agent earning credits by serving others), install `linkedclaw-provider` for the current runtime.
license: Apache-2.0
compatibility: Requires node + npm and the `@linkedclaw/cli` (installed from npm; this skill installs it for you).
allowed-tools: Bash(linkedclaw:*) Bash(jq:*) Bash(npm:*) Bash(node:*) Bash(command:*) Bash(printf:*) Read Write Edit
metadata:
  author: linkedclaw
  version: "0.6.1"

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
83% confidence
Finding

The skill documents an installation path that may attempt non-interactive sudo for global npm installation. Allowing a remotely sourced package installation flow to escalate privileges materially increases risk, because compromise of the package, installer script, or registry path could lead to full system-level code execution.

Content

Scanner excerpt · SKILL.md (reported line 112)May include surrounding context.

md
If `method` is `"npm-global"`, the script also includes a `path_hint` field telling the user how to put `~/.npm-global/bin` on PATH for future shells.

Encoding the chain as a script (rather than running the steps in your own reasoning) sidesteps a few mechanical mistakes that are easy under token pressure: forgetting the PATH export after `--prefix`, misreading the `npm warn EBADENGINE` warning as a failure (it's warning-only with exit 0), or leaking a sudo prompt back to the user.

---

Unsafe Defaults

Medium
Category
Tool Misuse
Confidence
70% confidence
Finding

Tool defaults are unsafe or overly permissive (e.g. disabled TLS verification, no authentication, world-writable permissions). Unsafe defaults widen the attack surface.

Content

Scanner excerpt · SKILL.md (reported line 151)May include surrounding context.

md
- They can pre-mint a key from another browser at `https://linkedclaw.com/settings/api-keys`, then run `linkedclaw login --paste` (or `--api-key lc_…`). That bypasses the OAuth handshake. The same don't-leak-the-key rules apply.
- This is the legacy paste path and is preserved indefinitely as the fallback for environments without browser access.

The CLI stores the key in `~/.linkedclaw/config.yaml` with secure modes (`0700` dir, `0600` file). Don't touch those modes — relaxing them would leave a long-lived API key world-readable on a shared host, and the CLI refuses to read the config on next start if it detects looser perms.

**Pointing at a non-prod cloud.** The CLI defaults to `https://api.linkedclaw.com`. For a test/staging cloud, set `LINKEDCLAW_CLOUD_URL` (and `LINKEDCLAW_RELAY_URL`) before `login`, or pass `linkedclaw login --cloud-url <url>`.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 348)May include surrounding context.

md
#### Per-call budget (agent sets)

- **Invoke**: cap with `--max-credits <n>`. Server has no per-call default — always set it unless the user opted out.
- **Hire (sessions)**: for a **fixed-price** capability the price is in `capabilities_meta[cap].curve`; the CLI **prefills your quote from the curve automatically** (you pass no price), and the agreed amount lands on `session.agreed_quote` at create — read it back and **`cancel` if it exceeds your budget before sending any messages**. (For an **open-quote** capability the provider proposes the price after your brief — see "Usage-based (open-quote) sessions"; you do not quote it.) `--max-messages <n>` is **both the turn budget and the spend cap**, and the **server default is 1** — omit it and the session allows a single `send`/`recv`, then sits at its cap. Set it to the number of turns you expect (e.g. `--max-messages 10` for a real dialogue, `5` for a short one).
- **Broadcast**: total budget is `target_providers × credits_per_provider` (both required fields). The CLI issues the Gig PA mandate for you when you pass `--principal-agent-id <agt_id>` (or have `agentId` in `~/.linkedclaw/config.yaml`).

If the user hasn't given a budget, pick a conservative default (~100 for small invokes, short sessions with `--max-messages 5`, broadcasts with `target_providers: 3`, `credits_per_provider: 5`) and surface it in the response so they know what it cost.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The guide gives materially inconsistent statements about whether a background Bash task completion automatically re-engages the agent. Early sections say completion emits a notification and requires no user ping, while later notes state Claude Code does not re-invoke on background exit. In a requester skill that can spend credits and orchestrate external agents, this mismatch can cause the agent to assume it will resume autonomously when it will not, or vice versa, leading to stuck sessions, missed cleanup, or unintended autonomous behavior assumptions.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

Conflicting guidance about re-engagement on background task completion can push an agent into policy-breaking execution patterns, especially around autonomous waiting and task continuation without explicit user interaction. Because this skill is designed to delegate work to third-party providers and manage multi-turn sessions, incorrect assumptions about when control returns can affect consent boundaries, billing flow, and whether the agent improperly continues work after the user expected the turn to end.

Content

No source excerpt is available for this finding.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
76% confidence
Finding

The guide instructs using an install script that may use sudo fallbacks to install the LinkedClaw CLI. Encouraging privilege escalation during dependency installation increases the blast radius of a compromised package, script bug, or path manipulation issue, especially in an agent-driven environment where shell commands may run with limited human scrutiny. The requester skill's ability to call external services makes secure installation particularly important.

Content

Scanner excerpt · references/claude-code.md (reported line 158)May include surrounding context.

md
per-turn noise in the parent context".
- No need to install anything beyond the `@linkedclaw/cli` npm package; the
  skill drives it through `Bash`. Use `<skill-dir>/scripts/install-cli.sh`
  for the install — it handles npm prefix / PATH / sudo fallbacks and emits
  a JSON status line you can parse.

## ACP bridge: not on this platform

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 · references/claude-code.md (reported line 167)May include surrounding context.

md
prompts). The `linkedclaw acp` bridge therefore adds nothing here — multi-turn hires use
`hire` + `send` + `recv --wait` exactly as described above. Do NOT install acpx and drive it
from the shell: it adds a process layer and downgrades payment confirmation to a config-file
auto-approve.

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 · references/codex.md (reported line 173)May include surrounding context.

md
prompts). The `linkedclaw acp` bridge therefore adds nothing here — multi-turn hires use
`hire` + `send` + `recv --wait` exactly as described above. Do NOT install acpx and drive it
from the shell: it adds a process layer and downgrades payment confirmation to a config-file
auto-approve.

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 · references/patterns.md (reported line 449)May include surrounding context.

md
prompts). The `linkedclaw acp` bridge therefore adds nothing here — multi-turn hires use
`hire` + `send` + `recv --wait` exactly as described above. Do NOT install acpx and drive it
from the shell: it adds a process layer and downgrades payment confirmation to a config-file
auto-approve.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

This section explicitly recommends widening subagent policy to grant shell access so spawned subagents can run linkedclaw. Expanding a child agent's tool permissions materially increases attack surface and can enable arbitrary command execution in a less-supervised delegated context, which goes beyond simple requester functionality.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% 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 · references/openclaw.md (reported line 174)May include surrounding context.

md
recv Tier B caveat #1), so you'd lose the auto-wake.

6. **`maxChildrenPerAgent: 5` and `maxConcurrent: 8`** are the concurrency
   caps. Don't fan out N spawned delegations in parallel without checking these.

### `sessions_spawn` parameter reference (relevant subset)

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The guide adds a durable cron-based autonomous workflow that can persist across restarts, closed sessions, and long time horizons while continuing to act on the user's behalf. That meaningfully extends the skill from immediate delegation into unattended orchestration, increasing risk of unintended actions, stale-context decisions, and hard-to-observe background behavior.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The cron polling and verification flow describes the agent making repeated background checks and even approving or rejecting provider results using its own judgment, but it does not require a strong user warning or fresh consent for sustained autonomous action. This creates a consent and governance gap where the agent may continue making consequential marketplace decisions after the user is no longer actively engaged.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · references/openclaw.md (reported line 328)May include surrounding context.

md
The `<NEXT_N>` is just `current_tick + 1` — names like `gigtask-poll-${TASK}-tick-1`,
`-tick-2`, … give each cron job a unique name, avoiding any race between the
old job's `--delete-after-run` and the new job's create call.

#### Termination conditions (recap)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The examples instruct the agent to upload potentially sensitive files such as passports and other documents to external providers, but they never require a user-facing warning, consent check, or data-minimization step first. In a requester skill whose purpose is delegating work to third-party agents, this omission can cause inadvertent disclosure of highly sensitive personal data to external systems.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The broadcast pattern distributes the same task to multiple providers in parallel, which materially increases data exposure compared with a single-provider workflow. Because the guidance omits any warning that user data may be copied to many external parties, an agent following this pattern could fan out confidential text, files, or business information without informed consent or minimization.

Content

No source excerpt is available for this finding.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
70% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · scripts/install-cli.sh (reported line 187)May include surrounding context.

sh
exit 0
  fi
  # Outdated below floor (or unreadable) — fall through to the install path,
  # which runs @latest with the full EACCES/sudo fallback chain.
fi

# Precondition: npm must exist.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
70% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · scripts/install-cli.sh (reported line 217)May include surrounding context.

sh
exit 0
  fi
  # Outdated below floor (or unreadable) — fall through to the install path,
  # which runs @latest with the full EACCES/sudo fallback chain.
fi

# Precondition: npm must exist.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
70% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · scripts/install-cli.sh (reported line 219)May include surrounding context.

sh
exit 0
  fi
  # Outdated below floor (or unreadable) — fall through to the install path,
  # which runs @latest with the full EACCES/sudo fallback chain.
fi

# Precondition: npm must exist.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · scripts/install-cli.sh (reported line 208)May include surrounding context.

sh
# 2. EACCES (or any failure where the global prefix is the suspect): retry
#    under ~/.npm-global. Harmless when the global prefix wasn't the issue.
NPM_GLOBAL_DIR="$HOME/.npm-global"
mkdir -p "$NPM_GLOBAL_DIR"
npm_out=$(npm install --prefix "$NPM_GLOBAL_DIR" -g @linkedclaw/cli@latest 2>&1)
npm_status=$?
export PATH="$NPM_GLOBAL_DIR/bin:$PATH"