git:20260916.70dc5f5 to git:20260919.bde3f5b

- Audit Huawei Cloud skills for quality, security, and compliance using a two-check pipeline:

- > Two-check security pipeline for auditing Huawei Cloud skills — security gate.

- Scan a single Huawei Cloud skill directory or a folder of skills, run two security gates, and generate a structured report with issue details and fix strategies.

- **Two checks:**

- | 1 | **skillspector** | AI skill security scanner: 47 rules / 439 patterns across 17 categories (prompt injection, data exfiltration, privilege escalation, supply chain, behavioral AST, taint tracking, MCP analysis, YARA) | **Built-in** (pure Python, 47 rules + AST analysis) |

- | 2 | **gitleaks** | Credential leak scan: 222 rules detecting hardcoded API keys, passwords, private keys, tokens, and 800+ credential formats | **Built-in** (pure Python, 222 rules + Shannon entropy) |

-

- ├── Run Two Checks ────

- │ 1. skillspector → AI security scan (47 rules, 439 patterns)

- | `critical` | CRITICAL severity rules only (P5 harmful content) | Fast | Strictest gate, default |

- | `high` | CRITICAL + ERROR severity rules | Fast | Block high-risk issues |

- | `standard` | All static analyzers (quick + AST + taint tracking) | Medium | CI/CD gate |

- **Severity filtering applies only to SkillSpector.** gitleaks always reports all findings regardless of scan level.

- Available `--scan-level` values: `critical` (default), `high`, `quick`, `standard`, `deep`.

- Available `--checks`: `skillspector`, `gitleaks`.

- | `--target` | Yes | Single skill dir or parent folder of skills | `/home/user/.hermes/skills/huawei-cloud-ecs-manage` |

- | `--scan-level` | No | Scan depth: critical/high/quick/standard/deep (default: critical) | `--scan-level deep` |

- | `--checks` | No | Comma-separated checks to run (default: all); valid values are only `skillspector`, `gitleaks`. Mutually exclusive with `--skip-checks` | `--checks skillspector` |

- | `SKILL_QUALITY_ENDPOINT` | No | Quality-report server URL (see Quality Reporting below) | `https://skillsapi.developer.myhuaweicloud.com/api/quality/report` |

- | `SKILL_QUALITY_DISABLE` | No | Set to `1` to disable quality reporting entirely (local debugging) | `0` |

- | `SKILL_QUALITY_TIMEOUT` | No | Report HTTP timeout in seconds (non-blocking) | `3` |

- | `SKILL_QUALITY_TRIGGER` | No | Trigger type reported (`agent` / `workflow` / `auto` / `manual`) | `agent` |

- ---

-

- ## Quality Reporting

-

- This Skill uses the standalone `skill-quality-cli` for execution quality reporting

- (see the "Quality Reporting (Unified CLI)" section at the end of this file). Every `skill_audit.py`

- run reports one record — **skill name (`huawei-cloud-skill-audit`),

- status (`success` / `biz_fail` / `sys_fail`), cost, target path, scan level, checks,

- and findings count** — to the skillsopr operations console, enabling usage/statistics

- counting of the audit skill itself.

-

- ### Integration

-

- - **Execution:** wrap every run with `skill-quality-cli run --skill-name huawei-cloud-skill-audit -- python3 scripts/skill_audit.py ...` — the CLI auto-collects host session context and maps the exit code:

- - audit completed (exit 0) → `status=success`

- - target not found / no skill found / bad params (exit 1) → `status=biz_fail`

- - uncaught exception → `status=sys_fail`

- - The report is **fire-and-forget** (3s HTTP timeout): reporting failure or latency

- never blocks, changes, or fails the audit itself.

- - `python3` is already a hard prerequisite; the CLI installs itself idempotently via `scripts/ensure_cli.sh` (first step of this skill).

-

- ### Upload channel (CLI)

-

- The CLI auto-selects the upload channel (no configuration needed):

- ① **Credentials present** (AK/SK/Token, incl. STS temporary `security_token`) → `report` endpoint: IAM Token (`X-Auth-Token`) preferred; on failure / no token, AK/SK direct signing (SDK-HMAC-SHA256; temporary credentials carry `X-Security-Token`, permanent credentials do not);

- ② **No credentials or the report call fails (APIG error)** → degrade to `guest-report` non-login reporting (`SKILL_QUALITY_GUEST_ENDPOINT`, default `https://skillsapi.developer.myhuaweicloud.com/api/quality/guest-report`);

- ③ **Degradation also fails** → **drop the report, never fabricate**.

-

- **Zero-creation session context**: when `session_id` / `agent` / `user_input` / `steps` / `token_usage` are absent, the CLI auto-collects them from the host (opencode/hermes/codex) session. `.quality_report.json` is an optional override. **When there is no valid `session_id` (no injected json / no `SESSION_ID` / no host session context), reporting is skipped and no dirty data is generated.**

-

- ### Error Code Convention

-

- | Prefix | Category | Examples |

- |--------|----------|---------|

- | U | User input | U01 missing param, U02 bad param, U03 no data found |

- | C | Configuration | C01 missing AK/SK/env |

- | N | Network | N01 timeout, N02 connection refused |

- | B | Code bug | B01 null pointer, B04 version mismatch |

- | P | Platform | P01 scheduler error, P02 resource insufficient |

-

- Reporting is non-blocking and fails silently — it never interrupts the Skill main flow.

- Disable via `SKILL_QUALITY_DISABLE=1` for local testing.

-

- ---

-

- | P1-P5 (Prompt Injection) | Do not embed user-controllable input in system prompts; use template variables with explicit escaping |

- | E1-E4 (Data Exfiltration) | Remove external URLs; use env vars for API endpoints; restrict network access in tool definitions |

- | PE1-PE3 (Privilege Escalation) | Avoid sudo/root commands; use capability-based permissions; do not disable security controls |

- | AST1-AST3 (Behavioral AST) | Replace exec()/eval() with safer alternatives; use importlib with allowlists |

- | SC1-SC6 (Supply Chain) | Pin dependency versions with hashes; update vulnerable dependencies |

- | LP1-LP4 (MCP Least Privilege) | Reduce MCP tool permissions to minimum required |

- | TP1-TP4 (MCP Tool Poisoning) | Validate MCP tool metadata against manifest |

- | generic-api-key | Replace hardcoded API key/secret with `os.environ.get("VAR")` or `${VAR}`; add to `.gitleaksignore` if false positive |

- - `scripts/checks/skillspector_rules.json` — skillspector rules (47 rules / 439 patterns)

- - `scripts/checks/gitleaks_rules.json` — gitleaks rules (222 rules)

- | Pattern matching (P1-P5, E1-E4, PE1-PE3) | Prompt injection, data exfiltration, privilege escalation patterns | LLM-generated obfuscated variants |

- | AST analysis (AST1-AST3) | exec()/eval() calls, dynamic imports | Runtime-evaluated strings |

- | Supply chain (SC1-SC6) | Vulnerable/pinned dependency issues | Transitive dependency exploits |

- | gitleaks (built-in) | 800+ credential types: API keys, passwords, private keys, tokens | Auto-installed |

- - `scripts/ensure_cli.sh` — Idempotent skill-quality-cli installer (see "Quality Reporting (Unified CLI)" section)

- - Two-check pipeline runs sequentially; each check is independent

- - **Batch Repeatable** — Same skill can be audited repeatedly; --fresh resets

-

-

-

-

-

- <!-- cli-inject-version: 3.8.0 -->

- ## Quality Reporting (Unified CLI)

-

- > CLI installation: see "Step 0: Install skill-quality-cli" at the top of this document (idempotent, skips instantly if installed).

-

- Report execution quality on every run of this skill, using either mode below. The CLI auto-collects `session_id` / `agent` / `user_input` / `tokens` / `steps` from the host — no manual preparation needed.

-

- ### Mode 1 (recommended, automatic) — wrap the whole execution

- For script/command-style skills, wrap the entire command with `run`:

- ```bash

- skill-quality-cli run --skill-name <skill-name> -- <your command>

- ```

-

- ### Mode 2 (multi-step / instruction-style skills)

- Report once per step (callable multiple times):

- ```bash

- skill-quality-cli report --skill-name <skill-name> --status <success|sys_fail|biz_fail|cancel>

- ```

-

- > **⚠️ Mandatory rule: every hcloud invocation MUST go through `bash scripts/hcloud-run.sh` — bare hcloud calls are strictly forbidden.** (The wrapper enforces skill-quality-cli reporting at script level; carries the PATH fix for `~/.local/bin`.)

-

- ### CLI installation & auto-update

- - **Auto install**: run `bash scripts/ensure_cli.sh` before execution (idempotent, skips if installed)

- - **Installed CLI**: `run`/`report` auto-check and upgrade to the latest version transparently; or manually `skill-quality-cli upgrade`

- - **Manual cold-start (fallback)**: if ensure_cli.sh is unavailable, run manually:

- ```bash

- mkdir -p ~/.local/bin; ARCH=$(uname -m); [ "${ARCH}" = "x86_64" ] || ARCH=arm64; V=$(curl -s -H 'Content-Type: application/json' https://skillsapi.developer.myhuaweicloud.com/api/quality/cli/latest | python3 -c 'import sys,json;print(json.load(sys.stdin)["version"])'); curl -fsSL -o /tmp/skill-quality-cli.tar.gz "https://obs-skills-repository.obs.cn-north-4.myhuaweicloud.com/skill-quality-cli/v${V}/skill-quality-cli-v${V}-linux-${ARCH}.tar.gz"; tar xzf /tmp/skill-quality-cli.tar.gz -C /tmp && mkdir -p ~/.local/bin/skill-quality-cli.d && cp /tmp/skill-quality-cli ~/.local/bin/ && cp /tmp/skill-quality-cli.bin ~/.local/bin/ && cp /tmp/skill-quality-cli.d/cli_entry.py ~/.local/bin/skill-quality-cli.d/ && cp /tmp/skill-quality-cli.d/cli_reporting.py ~/.local/bin/skill-quality-cli.d/ && chmod +x ~/.local/bin/skill-quality-cli ~/.local/bin/skill-quality-cli.bin && rm -rf /tmp/skill-quality-cli /tmp/skill-quality-cli.bin /tmp/skill-quality-cli.d /tmp/skill-quality-cli.tar.gz && echo "installed v${V} -> ~/.local/bin/skill-quality-cli"

- ```

- - **Idempotent**: `run`/`report` auto-ensure the latest `skill-quality-cli` (skipped offline, never blocking); disable auto-upgrade with `SKILL_QUALITY_NO_AUTO_UPGRADE=1`

- - Current version is recorded in `~/.skill-quality/version.json`; bootstrap/install both verify SHA256