Back to skill

Security audit

aidso-human-video-generator

Security checks for vulnerabilities and agentic risk

Overview

This video-generation skill needs review because unsafe authorization-file handling could let another local user redirect account access or overwrite files.

Prefer a version that validates authorization-file ownership and save destinations. If proceeding, set AIDSO_CREDENTIALS_FILE inside an existing private, user-owned directory to disable fallback searches. Submit only content you intend to share with AIDSO, and verify preview charges independently. Live authorization, billing, and service behavior were not tested.

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
  • 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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (29)

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 177)May include surrounding context.

md
- Treat authorization as runtime-only. Installation must not check whether a local binding file already exists.
- Installation summaries must present authorization as a first-use runtime step.
- Select and probe a writable local binding path before generating the authorization link.
- Default to `./.aidso/credentials.json`; fall back to the installed skill directory or a temporary runtime path if the working directory is not writable.
- Never ask the user to paste a key in chat.
- Never print or summarize the key, even masked, in install or status messages.
- Never ship, preload, or directly import a key through environment variables or package files.

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.yaml (reported line 25)May include surrounding context.

yaml
default_path: ./.aidso/credentials.json
    pending_path: ./.aidso/pending-authorization.json
    fallback_paths:
      - skill-directory/.aidso/credentials.json
      - tmp/aidso-human-video-generator/credentials.json
    permissions: "0600"
    created_by: scripts/auth.py

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 · references/api-spec.md (reported line 22)May include surrounding context.

md
Authorization is runtime-only. Installing this skill must not inspect local binding files; installation summaries should present authorization as a first-use runtime step.

Before generating an authorization link, select and probe a writable local binding path. The default path is `./.aidso/credentials.json` under the current working directory. If that path is not writable, try the installed skill directory, then a temporary runtime path. This prevents a successful authorization from being lost because the save location is blocked by an agent sandbox.

Run authorization in two separate commands:

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 · scripts/_aidso.py (reported line 25)May include surrounding context.

python
AUTH_LINK_TTL_SECONDS = 600
AUTH_LINK_CLOCK_SAFETY_MS = 1000
PENDING_AUTH_FILENAME = "pending-authorization.json"
DEFAULT_BINDING_RELATIVE_PATH = Path(".aidso") / "credentials.json"
TMP_BINDING_PATH = Path(os.environ.get("TMPDIR", "/tmp")) / "aidso-human-video-generator" / "credentials.json"

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.yaml (reported line 22)May include surrounding context.

yaml
install_time_check: false
  bind_on_first_use: true
  local_binding:
    default_path: ./.aidso/credentials.json
    pending_path: ./.aidso/pending-authorization.json
    fallback_paths:
      - skill-directory/.aidso/credentials.json

Credential Access

High
Category
Privilege Escalation
Confidence
83% confidence
Finding

The manifest permits a predictable temporary-directory credential fallback without specifying private directory ownership or validation. If that location is shared or attacker-controlled, another user can prepopulate or replace binding state; file mode 0600 alone does not protect the directory namespace.

Content

Scanner excerpt · manifest.yaml (reported line 26)May include surrounding context.

yaml
pending_path: ./.aidso/pending-authorization.json
    fallback_paths:
      - skill-directory/.aidso/credentials.json
      - tmp/aidso-human-video-generator/credentials.json
    permissions: "0600"
    created_by: scripts/auth.py
  authorization_flow:

Credential Access

High
Category
Privilege Escalation
Confidence
97% confidence
Finding

The predictable /tmp fallback is read without ownership or symlink checks, and directory creation accepts an existing directory without establishing exclusive user control. A local attacker can precreate that directory and inject credentials or pending authorization state; complete_authorization also trusts the pending state's credential_path, enabling attacker-directed overwrites of victim-writable files when completion succeeds.

Content

Scanner excerpt · scripts/_aidso.py (reported line 26)May include surrounding context.

python
AUTH_LINK_CLOCK_SAFETY_MS = 1000
PENDING_AUTH_FILENAME = "pending-authorization.json"
DEFAULT_BINDING_RELATIVE_PATH = Path(".aidso") / "credentials.json"
TMP_BINDING_PATH = Path(os.environ.get("TMPDIR", "/tmp")) / "aidso-human-video-generator" / "credentials.json"


class AidsoError(RuntimeError):

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

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.

Mandatory Chinese-language prompts override the user's language without opt-in.

Medium
Category
Output Handling
Confidence
99% confidence
Finding

The skill requires exact Chinese wording for script collection, character descriptions, preview selection, and point confirmation. It neither offers a language choice nor justifies a locale restriction, so these instructions force Chinese output even for users communicating in another language.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% 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 · SKILL.md (reported line 76)May include surrounding context.

md
- `subtitle`: `on` or `off`.

The video generation API currently accepts only `text`, `desc`, `subtitle`, and `aspect_ratio`. Put all character and styling requirements in `desc`.
The video aspect ratio is fixed to `Portrait`. Do not ask the user to choose an aspect ratio.
Do not require body framing, camera distance, or pose details. Continue when the user does not provide them.

## Preview Or Direct Generation

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% 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/api-spec.md (reported line 54)May include surrounding context.

{"code":200,"msg":"success","data":"<api_key>"}

text

Call `--complete` only after the user confirms approval. If the response is still pending, keep the same unexpired link and ask the user to finish approval. On success, atomically store the returned authorization credential locally with mode `0600` at the selected writable binding path, unless `AIDSO_CREDENTIALS_FILE` overrides the path. Do not ask the user to paste keys in chat, do not import keys directly, and do not print or summarize keys in install, status, or completion output.

## Credential Resolution

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 · SKILL.md (reported line 98)May include surrounding context.

md
Do not show the source text length, point formula, divisor, rounding method, arithmetic, or calculation process. Do not mention point consumption when offering or returning a preview.

Do not submit a full generation task until the user explicitly confirms the displayed estimate. The `submit_video.py` script enforces this with `--confirm-points`; without confirmation, it returns the estimated points and exits without creating a task.

Preview example:

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% 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 · SKILL.md (reported line 178)May include surrounding context.

md
- Installation summaries must present authorization as a first-use runtime step.
- Select and probe a writable local binding path before generating the authorization link.
- Default to `./.aidso/credentials.json`; fall back to the installed skill directory or a temporary runtime path if the working directory is not writable.
- Never ask the user to paste a key in chat.
- Never print or summarize the key, even masked, in install or status messages.
- Never ship, preload, or directly import a key through environment variables or package files.
- Read the local binding file only after it was created by the authorization flow.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% 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 · manifest.yaml (reported line 45)May include surrounding context.

yaml
- Do not fetch or validate the JavaScript authorization page with a plain HTTP request.
    - Query the pending code only after the user confirms approval.
    - Generate a new link instead of reusing an expired link.
    - Do not ask users to paste keys into chat.
    - Do not print or summarize keys, even masked, in install or status output.
    - Do not support direct key import through environment variables or package files.
    - Store the returned credential locally only after the user approves the AIDSO authorization link.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 · manifest.yaml (reported line 54)May include surrounding context.

yaml
base_url: https://api.aidso.com
  auth_generate_code:
    method: GET
    url: https://api.aidso.com/openapi/skills/generate_code
  auth_get_by_code:
    method: GET
    url: https://api.aidso.com/openapi/skills/get_by_code

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 · manifest.yaml (reported line 57)May include surrounding context.

yaml
url: https://api.aidso.com/openapi/skills/generate_code
  auth_get_by_code:
    method: GET
    url: https://api.aidso.com/openapi/skills/get_by_code
    query:
      - code
  generate_video:

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 · manifest.yaml (reported line 62)May include surrounding context.

yaml
- code
  generate_video:
    method: POST
    url: https://api.aidso.com/openapi/skills/gen_digital_human_video
    body:
      - text
      - desc

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 · manifest.yaml (reported line 89)May include surrounding context.

yaml
- Content-Type: application/json
  get_video_result:
    method: GET
    url: https://api.aidso.com/openapi/skills/get_digital_human_video_result
    headers:
      - x-api-key
  submit_text_extract_task:

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 · manifest.yaml (reported line 102)May include surrounding context.

yaml
- Content-Type: application/json
  get_text_extract_result:
    method: GET
    url: https://api.aidso.com/openapi/skills/get_text_extract_result
    query:
      - taskId
    headers:

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

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

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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/api-spec.md (reported line 13)May include surrounding context.

User-facing prompts are fixed in Chinese without a language choice or documented locale requirement.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The configured description prompt and submission-confirmation message prescribe Chinese text for user interactions. The manifest provides no language opt-in, localization mechanism, or region-specific justification for imposing this language.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 · manifest.yaml (reported line 94)May include surrounding context.

yaml
- x-api-key
  submit_text_extract_task:
    method: POST
    url: https://api.aidso.com/openapi/skills/submit_text_extract_task
    body:
      - url
    headers:

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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/api-spec.md (reported line 8)May include surrounding context.

Use the production endpoints for all account binding and generation operations:

http
GET https://api.aidso.com/openapi/skills/generate_code
GET https://api.aidso.com/openapi/skills/get_by_code?code=<code>
POST https://api.aidso.com/openapi/skills/submit_text_extract_task
GET  https://api.aidso.com/openapi/skills/get_text_extract_result?taskId=<taskId>

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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/api-spec.md (reported line 10)May include surrounding context.

http
GET https://api.aidso.com/openapi/skills/generate_code
GET https://api.aidso.com/openapi/skills/get_by_code?code=<code>
POST https://api.aidso.com/openapi/skills/submit_text_extract_task
GET  https://api.aidso.com/openapi/skills/get_text_extract_result?taskId=<taskId>
POST https://api.aidso.com/openapi/skills/gen_digital_human_video
GET  https://api.aidso.com/openapi/skills/get_digital_human_video_result?taskId=<taskId>