BB huawei-cloud-skill-audit
Audit Huawei Cloud skills for quality, security, and compliance using a two-check pipeline: skillspector (AI security) and gitleaks (credential leak). Generates structured reports with issue details and fix strategies. Triggers include: "审计技能","技能审计","检查技能质量","扫描技能问题","技能安全审计", "audit skill","check skill quality","scan skills for issues","skill audit", "华为云技能审计","技能合规检查","skill gate","质量门禁","技能检查", "audit huawei cloud skill","verify skill compliance","技能质量检查","跑审计","安全扫描".
Audit Huawei Cloud skills for quality, security, and compliance using a two-check pipeline: skillspector (AI security) and gitleaks (credential leak).
As a process B 70/100 · Nearly there — weak spots: when it triggers, running it twice, progress reporting
AnalyzerSecurityQuality controlInfrastructuretype and topics are labelled automatically from the skill text
How to improve
- Say in the description WHEN to use the skill ("use when…", example requests): that is the agent's main cue.
For the model run — optional
- Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 16
✓ No critical or high findings
Medium and low: 16
-
low Risky intent
intent-offensive-securityreferences/acceptance-criteria.md:26Offensive-security / dual-use content (legitimate for authorised testing; review intended use)- No privilege escalation patterns (PE1-PE3)
-
low Risky intent
intent-offensive-securityreferences/gitcode-security-scanner.md:15Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| **Risk domain** | AI safety (reverse shell, command injection, prompt injection, eval/exec) | InfoSec (credential leak, SQL injection, path traversal, debug leakage) |
-
low Risky intent
intent-offensive-securityreferences/iam-policies.md:34Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| No sudo/root commands | skillspector | No privilege escalation patterns |
-
low Risky intent
intent-offensive-securityreferences/security-audit-guide.md:10Offensive-security / dual-use content (legitimate for authorised testing; review intended use)- Privilege escalation (PE1-PE3)
-
low Risky intent
intent-offensive-securityreferences/security-audit-guide.md:22Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| Privilege escalation | Avoid sudo/root; use capability-based permissions |
-
low Risky intent
intent-offensive-securityreferences/security-audit-guide.md:24Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (detector / deny-list definition)| YARA matches | Remove reverse shell/webshell patterns |
detector -
low Secrets in code
secret-high-entropy-tokenscripts/checks/gitleaks_rules.json:235High-entropy token-like string (may be an id, hash or a credential) (detector / deny-list definition)"regex": "bedr…29t",
detector -
low Exfiltration
intent-browser-credential-storescripts/checks/skillspector_rules.json:599Accesses a browser credential / cookie store (detector / deny-list definition; test fixture / example file)"regex": "(?:Chrome|Firefox|Safari)/.*?(?:Cookies|Login Data|key4\\.db)",
detectorfixture -
low Dangerous commands
cmd-background-processscripts/checks/skillspector_rules.json:1421Starts a background / autostarted process (detector / deny-list definition; string literal in code, not executed; test fixture / example file)"regex": "(?:nohup|disown|setsid)\\s+",
detectorcode literalfixture -
low Risky intent
intent-offensive-securitySKILL.md:28Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (detector / deny-list definition)| 1 | **skillspector** | AI skill security scanner: 47 rules / 439 patterns across 17 categories (prompt injection, data exfiltration, privilege escalation, supply chain, behavioral AST, taint trackin
detector -
low Risky intent
intent-offensive-securitySKILL.md:219Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| PE1-PE3 (Privilege Escalation) | Avoid sudo/root commands; use capability-based permissions; do not disable security controls |
-
low Risky intent
intent-offensive-securitySKILL.md:221Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| YR1-YR4 (YARA) | Remove reverse shell/webshell patterns; move server functionality to separate controlled service |
-
low Risky intent
intent-offensive-securitySKILL.md:289Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (detector / deny-list definition)| Pattern matching (P1-P5, E1-E4, PE1-PE3) | Prompt injection, data exfiltration, privilege escalation patterns | LLM-generated obfuscated variants |
detector -
low Risky intent
intent-offensive-securitySKILL.md:291Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (detector / deny-list definition)| YARA rules (YR1-YR4) | Reverse shell, webshell patterns | Encoded/obfuscated payloads |
detector
A further 2 matches are quotations in this security skill's documentation and are not counted as findings.
Files scanned: 23. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
- warning
description-no-whendescription does not say WHEN to use the skill (no "use when")
Process rating: all ten parameters 70/100
- 0Progress reporting. Says nothing while it works
- 20When it triggers. No condition that starts the skill
- 30Running it twice. 5 mutating operations with no state check
- 50Failures and branches. 0 branches, has a failure section
- 60Tools and files. Uses tools (bash, python) that frontmatter does not declare
- 70Inputs and preconditions. Inputs and preconditions are listed
- 100Steps. 47 steps
- 100Result and completion. Output format and completion criterion are stated
- 100Consistency. Name and required fields are in place
- 100Execution cost. Instruction body is 3942 tokens
- medium Safety rules and hard prohibitions inside a skill: they belong in the system prompt, here they protect nothing
- low 21 top-level sections: this looks like several domains in one skill
- low The response is described with custom markup (3 tags): a typed call is more reliable
Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.
Quality signals
- +4Description does not say when NOT to use the skill (false activations)
- -4Absolute local paths (C:\Users, /home/…): not portable
- -33 of 5 scripts are never mentioned in SKILL.md
- +1No license
- +2Single-language instructions
- +5Description quotes 13 example trigger phrases
- +3Description length 477: enough signal without eating the budget
- +4Structure: 36 headings
- +3Step-by-step instructions: 47 items
- +3Output format is stated explicitly
- +4Has examples (11 code blocks)
- +4Reference files are cited in the instructions (5 of 5)
Quality base 70; lint remarks subtract, signals add up to 100. Result: 77.
External checks
ClawHub: suspicious
The skill performs the advertised local audit, but it also automatically reports metadata and probes local agent sessions and cloud credentials in ways users are not fully told about.
LLM: suspicious (high) · 10 Sept 2026