跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
浏览器与自动化

rl-protect-edit-profile

Creates or modifies rl-protect scanning profiles (rl-profile JSON). Use when the user wants to configure scan policy, set assessment severity levels, add governance allow/block rules, or apply policy overrides.

reversinglabs · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add reversinglabs/rl-protect-skills --skill rl-protect-edit-profile

技能说明


name: rl-protect-edit-profile description: Creates or modifies rl-protect scanning profiles (rl-profile JSON). Use when the user wants to configure scan policy, set assessment severity levels, add governance allow/block rules, or apply policy overrides.

Profile editing (rl-protect)

Use this skill when the user wants to create a new profile or change an existing one. Profiles are JSON files that control how rl-protect evaluates packages — what counts as a failure, which packages are always allowed or blocked, and which policy rules are overridden.

Profile location

The profile is loaded from a file passed to rl-protect scan via --profile={path}. If the user does not specify a path, rl-protect will use its default profile. Read the file before making any changes if it already exists.


Schema reference

rl-profile
  .schema                          — always 1, do not change
  .name                            — human-readable profile name
  .info                            — description string (free text)
  .timestamp                       — ISO-8601, set to current time on every save
  .configuration.policy
    .min_catalogue                 — minimum required catalogue level (integer, 1–10)
    .rl_scan_level                 — scan depth level (integer, 1–10)
    .rl_auto_approval              — auto-approve packages that pass (boolean)
    .assessments{}                 — one entry per check category (see below)
    .governance.community          — age gates and allow/block pattern lists
    .overrides{}                   — keyed by policy rule ID (SQ-prefixed)

Assessment configuration

Each assessment can be ignored entirely or configured with per-detection-type severity levels.

Valid severity values: pass | warning | fail

assessments.secrets
  .ignored                         — true suppresses the entire check

assessments.licenses
  .ignored                         — true suppresses the entire check

assessments.vulnerabilities
  .ignored
  .detections
    .mandate                       — CVEs under active government mandate
    .exploit                       — CVEs with known exploits
    .malware                       — CVEs used in malware campaigns
    .critical                      — CVEs with CVSS ≥ 9.0

assessments.hardening
  .ignored

assessments.tampering
  .ignored
  .ml_hunting                      — ML-based anomaly detections

assessments.malware
  .ignored
  .rl_analyst                      — findings from RL analyst vetting
  .rl_scanner                      — findings from RL scanner
  .suspicious                      — suspicious (unconfirmed) detections
  .dependency
    .develop                       — malicious dependency in dev context
    .release                       — malicious dependency in release context
  .detections                      — affects specified malware types
    .adware
    .riskware
    .protestware
    .spam

Governance configuration

Governance rules control package-level allow and block decisions independently of assessment findings.

governance.community
  .min_package_age                 — minimum days since first package release (integer)
  .min_version_age                 — minimum days since this version was published (integer)
  .allow[]                         — packages always approved regardless of findings
    .pattern                       — PURL pattern (exact or wildcard, e.g. pkg:npm/lodash@*)
    .audit.author                  — who added the rule
    .audit.timestamp               — when the rule was added
    .audit.reason                  — stated justification
  .block[]                         — packages always rejected regardless of findings
    .pattern                       — PURL pattern
    .audit.author                  — who added the rule
    .audit.timestamp               — when the rule was added
    .audit.reason                  — stated justification

PURL pattern matching:

  • Exact version: pkg:npm/lodash@4.17.21
  • Wildcard version range: pkg:npm/ua-parser-js@0.7.*
  • All package versions: pkg:gem/rack@*
  • Specific artifact: pkg:pypi/flask@3.1.2?artifact=flask-3.1.2-py3-none-any.whl

Policy overrides

Overrides change the effective status of specific policy rule violations by their rule ID.

overrides.{SQxxxxx | THxxxxx}
  .enabled                         — true to activate the override
  .blocker                         — the status to apply instead: pass | fail
  .apply_to[]                      — scope: "organization" | "group" | "any"
  .audit.author
  .audit.timestamp                 — must use current timestamp in ISO-8601 format
  .audit.reason

Setting blocker to pass means the rule never blocks a scan. Setting it to fail elevates a warning to a failure.


Tasks


Create a new profile

Trigger: user wants to create a profile from scratch, or no profile file exists.

Steps:

  1. Ask the user for a profile name and any initial settings they want to configure.
  2. Generate a valid profile JSON with schema: 1 and the current timestamp.
  3. Set sensible defaults for any fields the user did not specify (see defaults below).
  4. Write the file and confirm the path.

Default values for new profiles:

FieldDefault
min_catalogue5
rl_scan_level5
rl_auto_approvalfalse
All assessment .ignoredfalse
vulnerabilities.detections.mandatefail
vulnerabilities.detections.exploitfail
vulnerabilities.detections.malwarefail
vulnerabilities.detections.criticalfail
malware.rl_analystfail
malware.rl_scannerfail
malware.suspiciouswarning
malware.dependency.releasefail
malware.dependency.developwarning
min_package_age90
min_version_age3

Configure assessment severity

Trigger: user wants to ch

阅读完整来源 ↗