Security checks for vulnerabilities and agentic risk
Overview
This precheck skill is not destructive, but it should be reviewed because it overstates link checking and can print detected secrets into logs.
Install only if you are comfortable treating it as a local syntax and pattern checker, not a real link availability verifier. Avoid running check or all on files that may contain live credentials in shared terminals, CI, or logged agent sessions unless the output is protected or the tool is patched to redact secret matches.
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
Warning
Location
scripts/verify.py:77
Finding
Sensitive values may be exposed in terminal output and logs
Content
View full analysis
Vulnerability Details
File Location: scripts/verify.py, lines 77–82 and 179–181 Vulnerability Type: Sensitive-data exposure through diagnostic output Risk Level: Medium
Vulnerable Code
python
issues.append({
'severity': 'CRITICAL' if '密钥' in desc or 'Key' in desc or '私钥' in desc else 'WARNING',
'line': i,
'description': desc,
'match': line.strip()[:80],
})
python
if issue['match']:
print(f" └─ {issue['match'][:70]}")
Technical Analysis
The secret scanner stores up to 80 characters from the complete source line containing a detected credential. The all and check command paths subsequently print up to 70 characters from that stored line.
Because the output is not redacted, it may reproduce API keys, passwords, access keys, tokens, or private-key content. This conflicts with the scanner's security purpose: content identified as sensitive is copied into another disclosure channel. Terminal output can be retained in shell transcripts, CI/CD job logs, agent execution records, monitoring systems, or support artifacts.
The dedicated secrets command does not print the match value, but both all and check invoke run_all_checks(), which does.
Attack Path
A scanned file contains a credential on a line matching one of the configured secret patterns.
A user or automated pipeline runs:
bash
python3 scripts/verify.py all sensitive-file.md
or:
bash
python3 scripts/verify.py check sensitive-file.md
check_secrets() copies the first 80 characters of the affected source line into the issue's match field.
run_all_checks() prints up to 70 characters of that field without redacting the credential.
The exposed value may be retained by terminal capture, CI logging, monitoring, or agent-output storage and become accessible to parties or systems that could not read the original file.
Impact Assessment
This issue does not directly gran
...[truncated 574 chars]
Remediation
View remediation
Remediation Suggestions
Do not store or print the original source line for secret findings.
Report only the file name, line number, severity, and detected secret type.
Replace the match value with a constant marker such as [REDACTED], or omit the field entirely.
If limited context is operationally necessary, redact every matched credential value before retaining or displaying the line.
Keep unredacted values out of exceptions, debug logs, structured reports, and telemetry.
Add automated tests using canary credentials and assert that no complete or partial credential appears in output from all, check, or secrets.
A safer implementation would be:
python
issues.append({
'severity': 'CRITICAL' if '密钥' in desc or 'Key' in desc or '私钥' in desc else 'WARNING',
'line': i,
'description': desc,
'match': '[REDACTED]',
})
Preferably, remove the match field entirely for secret findings.
The skill advertises security-relevant checks such as link availability validation and implementation with requests, but the documented behavior indicates those protections are not actually performed. This can create a false sense of safety, leading users to publish content under the mistaken belief that links were validated and broader verification occurred, which is especially risky for a pre-publication security-checking tool.
Content
No source excerpt is available for this finding.
Undeclared Tool Scope
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.
Natural-Language Policy Violations
Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding
The natural-language description forces a specific language/locale for users reading the skill metadata. The file does not offer an opt-in, alternative language, or justification that this is a region-specific skill, which matches the language-policy violation criteria.
Content
No source excerpt is available for this finding.
Description-Behavior Mismatch
Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding
The skill metadata claims automatic link availability checking, but the implementation only parses Markdown links and emits an informational message for HTTP URLs without performing any network reachability test. In a pre-publication verification tool, this creates a false sense of assurance: users may believe broken or maliciously swapped links were validated when they were not, causing security and quality checks to be silently bypassed.
Content
No source excerpt is available for this finding.
Intent-Code Divergence
Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding
The docstring says the function checks Markdown link format, which contradicts the broader product description claiming link reachability validation. This inconsistency is dangerous because operators and downstream agents may trust the stronger claim in the manifest and skip independent validation, leading to missed broken links or unsafe destinations.