Back to skill

Security audit

Finder 达人搜索

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent Finder creator-search helper, but it normalizes users pasting API keys into chat and stores them locally in plaintext without adequate safeguards.

Review before installing. Use this skill only if you are comfortable with Finder API access from the agent. Do not paste long-lived API keys into chat; prefer configuring credentials yourself through a secure local mechanism, restrict file permissions if ~/.finder/config.json is used, and revoke or rotate the key if it was shared in conversation.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T09 · Insecure Skill Coding Practices

Error
Location
references/config.md:44
Finding

Unsafe API-Key Interpolation and Plaintext Credential Storage

Content
View full analysis
"$HOME/.finder/config.json" <", "last_project_id": null } EOF ``` The documented Windows PowerShell implementation similarly places the value in executable PowerShell source: ```powershell $finderDir = Join-Path $HOME ".finder" $configPath = Join-Path $finderDir "config.json" New-Item -ItemType Directory -Force -Path $finderDir | Out-Null $config = @{ base_url = "https://finder.optell.com" api_key = "" last_project_id = $null } | ConvertTo-Json Set-Content -Path $configPath -Value $config -Encoding UTF8 ``` ### Technical Analysis The Skill expects an agent to replace the API-key placeholder with user-controlled content and execute the resulting Bash or PowerShell command. The value is neither validated against an expected API-key format nor passed through a non-executable data channel. In the Bash example, the here-document delimiter is unquoted. Consequently, shell expansions such as command substitution are processed inside the here-document. If an untrusted value containing shell syntax such as `$(command)` is inserted into the generated command, that command may be evaluated while the configuration file is created. A value containing a newline and the `EOF` delimiter could also terminate the i ...[truncated 2660 chars]
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (26)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The documentation encourages users to paste sensitive API credentials into the conversation without warning about confidentiality, retention, or least-privilege concerns. This creates a direct secret-handling anti-pattern that can expose credentials to logs, transcripts, plugins, or other unintended storage layers.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill instructs the assistant to have the user paste an access key into chat so the assistant can continue setup and usage. This is dangerous because it turns the assistant into a credential collection point and increases risk of disclosure through message history, telemetry, or accidental reuse.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill explicitly tells the assistant to ask users to send API keys in chat and then use them to create local configuration files. Soliciting secrets through conversation is unsafe because chat channels may be logged, retained, or exposed, and storing the key locally creates persistent credential risk beyond the stated creator-search function.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

Telling users they can directly send their access key in chat for the assistant to create config files couples secret exfiltration with persistence. This materially raises the attack surface because the key may be captured both in transcripts and on disk, and the behavior is not essential to the skill's narrow search purpose.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

This scenario explicitly operationalizes the unsafe flow: user pastes API key, assistant saves it to a local file, and then continues actions using that credential. That is a concrete secret-handling vulnerability because it institutionalizes collection, storage, and use of sensitive credentials in a conversational tool path.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly directs the agent to accept an API key pasted into chat and write it into a local config file, but provides no warning about the sensitivity of the credential, risks of chat disclosure, or file-permission/security requirements. This normalizes handling secrets through conversational channels and automatic persistence, which can expose credentials through logs, transcript retention, or insecure local storage.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

This instruction tells the agent to persist a user-provided API key from chat into a local credential file, which treats a sensitive secret shared in conversation as normal operational input. That increases the chance of credential leakage via chat logs, agent memory, debugging traces, or insecure filesystem permissions on the created config file.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The workflow goes beyond mentioning storage and specifically instructs the agent to automatically create the credential file when a key is pasted into chat. This creates a direct secret-ingestion and persistence path that may occur without adequate consent, secure handling guarantees, or safeguards against retention in system logs and transcripts.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The instruction 用中文,简短直接 imposes a specific language for all interactions. This is a language/locale policy issue because the skill does not offer user opt-in or indicate that the skill is restricted to Chinese-speaking users for a justified regional reason.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

Telling the agent not to ask for a previously supplied key again implies the system may retain and reuse a credential originally disclosed in chat. That expectation encourages persistent handling of secrets across turns and increases the risk of accidental disclosure, unauthorized reuse, or unclear retention boundaries.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The example response reinforces that the agent has written the user's access key to a local config file and that the user need not re-enter it, normalizing secret persistence from chat as a recommended behavior. Examples are powerful operational guidance, so this increases the likelihood that downstream agents will actually collect, store, and retain credentials insecurely.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The display name, description, and default prompt are all written exclusively in Chinese, which implies a fixed language/locale for the skill experience without any opt-in or alternative. Under the policy, language constraints should either offer user choice or be clearly documented as justified for a region-specific tool.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
82% confidence
Finding

The entire skill documentation, including recommended user-facing phrases, is written exclusively in Chinese and does not offer an opt-in or alternative language. This can violate language/locale policy when a skill implicitly constrains interactions to one language without user choice or documented justification.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The documentation instructs the agent to accept a user's Finder API key and persist it locally, which expands the skill from simple creator search into credential handling and local filesystem modification. That broader capability increases the blast radius if the agent is misused, the host is compromised, or the user did not intend long-term plaintext storage of the credential.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The documentation explicitly encourages the user to paste the API key to the agent so it can be saved, but it provides no privacy or security warning about sharing a secret with an automated system. This can lead users to disclose long-lived credentials without understanding retention, logging, or downstream exposure risks.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The markdown provides exact commands for writing the API key into a plaintext local configuration file, but gives no warning about local credential storage risk. This can normalize insecure secret handling and increase the chance of credential theft from local compromise, backups, or accidental file disclosure.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The documented commands store the API key in a local JSON file in plaintext, which is unnecessary for a narrowly scoped search skill and creates a credential exposure risk. Any local process, backup system, shell history leakage, or shared account access could reveal the key and allow unauthorized API use.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill documentation goes beyond influencer search by directing automatic remote project creation in the user's Finder account. This is scope expansion that enables state-changing actions on an external service, which can surprise users, create unwanted resources, and normalize broader account operations than the manifest suggests.

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 · references/config.md (reported line 134)May include surrounding context.

md
print(json.load(open(path)).get("api_key", ""))
PY
)
curl -X POST 'https://finder.optell.com/api/project/create' \
  -H "Authorization: Bearer $KEY" \
  -H 'Content-Type: application/json' \
  -d '{

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The documented behavior expands beyond creator search into collecting and persisting a user's API key in local configuration. This broadens the skill's effective privileges and data handling responsibilities, increasing the chance of credential misuse or unintended retention without clear necessity or safeguards.

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 · references/config.md (reported line 172)May include surrounding context.

md
print(json.load(open(path)).get("api_key", ""))
PY
)
curl -X POST 'https://finder.optell.com/api/creator/search' \
  -H "Authorization: Bearer $KEY" \
  -H 'Content-Type: application/json' \
  -d '{

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 · references/examples.md (reported line 37)May include surrounding context.

md
print(json.load(open(path)).get("api_key", ""))
PY
)
curl -X POST 'https://finder.optell.com/api/creator/search' \
  -H "Authorization: Bearer $KEY" \
  -H 'Content-Type: application/json' \
  -d '{

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The example normalizes writing a long-lived credential to ~/.finder/config.json without any notice about local file permissions, device sharing, or secret lifecycle. While local storage can be legitimate, presenting it without safeguards encourages insecure handling of sensitive material.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

The recommended confirmation message reassures the user that the assistant has stored the secret and that they need not enter it again, reinforcing insecure secret persistence as normal behavior. This can reduce user caution and encourage repeated use of unsafe credential-handling patterns.

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 · references/examples.md (reported line 136)May include surrounding context.

md
print(json.load(open(path)).get("api_key", ""))
PY
)
curl -X POST 'https://finder.optell.com/api/creator/search' \
  -H "Authorization: Bearer $KEY" \
  -H 'Content-Type: application/json' \
  -d '{