Back to skill

Security audit

每日罗摩衍那 Daily Ramayana

Security checks for vulnerabilities and agentic risk

Overview

The skill’s daily reading workflow fits its purpose, with manageable caveats around online narration and separately supplied dependencies.

Narration sends its script to Microsoft’s online speech service, as documented by [edge-tts](https://github.com/rany2/edge-tts); keep private information out of it or set NO_TTS=1. Pin the speech dependency and review the separately supplied archive helpers before use. The package also needs the missing book text. Inspection did not substantiate the scanner’s language-mismatch, excessive-permission, or credential-flow concerns.

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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (7)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

Progress management is consistent with the description, but explicitly selecting English conflicts with the advertised Chinese presentation. The description permits an English source excerpt, not an English overall language setting. The script implementation is not supplied, so the precise effect of this setting cannot be verified.

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, file_write.

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.

The narration workflow sends generated reading text to an external speech service without disclosing that transmission.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

The narration workflow transmits the reading script to an external speech service without clearly telling the user that the text leaves the local environment. The normal payload is generated commentary on public-domain literature, substantially limiting confidentiality impact, but naming edge-tts does not adequately disclose the data transfer.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'uvx edge-tts'.

Medium
Category
MCP Rug Pull
Confidence
96% confidence
Finding

The fallback invokes an unpinned Python package through uvx, potentially downloading and executing a changing dependency with the agent's permissions. This is a supply-chain exposure, although edge-tts is a CLI package rather than an MCP server.

Content

No source excerpt is available for this finding.

Tainted flow: 'PROGRESS_FILE' from os.environ.get (line 31, credential/environment) → pathlib.Path.write_text (file write)

Medium
Category
Data Flow
Confidence
65% confidence
Finding

Data from a source is assigned to a variable that is later passed to a sink, creating a variable-mediated taint flow.

Content

Scanner excerpt · scripts/progress.py (reported line 73)May include surrounding context.

python
def save(data):
    PROGRESS_FILE.parent.mkdir(parents=True, exist_ok=True)
    PROGRESS_FILE.write_text(json.dumps(data, ensure_ascii=False, indent=2), encoding="utf-8")


def book_body(book):

MCP server referenced without pinned version: 'uvx edge-tts'.

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

The script executes edge-tts through uvx without specifying a reviewed version, leaving executed code dependent on package resolution and future upstream releases. A compromised resolved dependency could execute with the script's permissions; this is a package supply-chain issue, not an unpinned MCP server.

Content

No source excerpt is available for this finding.

Input text is sent to an external TTS service without disclosing the transmission.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

Passing the input file to edge-tts sends its text to an external speech service, while the script provides no explicit privacy notice or consent gate. Intended inputs are non-sensitive literary narration, reducing the normal impact, but any private text included in that file would also be transmitted.

Content

No source excerpt is available for this finding.