Rl Protect Edit Profile

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:

| Field | Default | |---|---| | min_catalogue | 5 | | rl_scan_level | 5 | | rl_auto_approval | false | | All assessment .ignored | false | | vulnerabilities.detections.mandate | fail | | vulnerabilities.detections.exploit | fail | | vulnerabilities.detections.malware | fail | | vulnerabilities.detections.critical | fail | | malware.rl_analyst | fail | | malware.rl_scanner | fail | | malware.suspicious | warning | | malware.dependency.release | fail | | malware.dependency.develop | warning | | min_package_age | 90 | | min_version_age | 3 |

---

Configure assessment severity

Trigger: user wants to change what counts as a failure or warning for a specific check, or wants to ignore a check entirely.

Steps:

1. Identify which assessment and detection type the user wants to change.
2. Confirm the new severity value (pass, warning, or fail).
3. Update only the affected field. Do not alter other assessments.
4. Update .timestamp to the current time in ISO-8601 format.
5. Write the file and show a summary of what changed.

Output summary format:

### Profile updated — {profile name}

| Assessment | Detection | Previous | New |
|---|---|---|---|
| {assessment name} | {detection type or "ignored"} | {old value} | {new value} |

---

Add a governance rule

Trigger: user wants to always allow or always block a specific package or version range.

Steps:

1. Ask for: the PURL pattern, whether it is an allow or block rule, and a reason.
2. Check whether a rule for the same pattern already exists. If so, warn the user and ask whether to replace it.
3. Populate audit.author from context if known, otherwise ask. Prefer emails. Set audit.timestamp to the current time in ISO-8601 format.
4. Append the rule to the appropriate list (allow[] or block[]).
5. Update .timestamp and write the file.

Output summary format:

### Governance rule added — {profile name}

| Field | Value |
|---|---|
| Action | {Allow / Block} |
| Pattern | {purl pattern} |
| Reason | {reason} |
| Author | {author} |

---

Remove a governance rule

Trigger: user wants to remove an existing allow or block rule.

Steps:

1. List all current rules in the relevant list and ask the user to confirm which to remove.
2. Remove the matching entry by pattern.
3. Update .timestamp and write the file.

---

Add or update a policy override

Trigger: user wants to suppress or downgrade a specific policy rule violation by its SQ rule ID.

Steps:

1. Ask for: the SQ rule ID, the target status (pass or warning), the scope (organization, group, or any), and a reason.
2. If an override for that rule ID already exists, show the current values and ask the user to confirm the update.
3. Set enabled: true, populate the audit fields, and update .timestamp in ISO-8601 format.
4. Write the file and confirm.

Output summary format:

### Policy override saved — {profile name}

| Field | Value |
|---|---|
| Rule | {SQxxxxx} |
| Overridden to | {pass / warning} |
| Scope | {organization / group / any} |
| Reason | {reason} |
| Author | {author} |

---

Show current profile

Trigger: user wants to review what is currently configured in the profile.

Read the profile file and present it in this format:

### Profile — {name}

**Scan settings**
| Setting | Value |
|---|---|
| Min catalogue level | {min_catalogue} |
| Scan level | {rl_scan_level} |
| Auto-approval | {Yes / No} |

**Assessment configuration**
| Assessment | Ignored | Detection severities |
|---|---|---|
| Secrets | {Yes/No} | — |
| Licenses | {Yes/No} | — |
| Vulnerabilities | {Yes/No} | mandate: {v} · exploit: {v} · malware: {v} · critical: {v} |
| Hardening | {Yes/No} | — |
| Tampering | {Yes/No} | ml_hunting: {v} |
| Malware | {Yes/No} | analyst: {v} · scanner: {v} · suspicious: {v} · dep/release: {v} · dep/develop: {v} |

**Governance**
- Min package age: {N} days · Min version age: {N} days
- Allow rules: {N}  ·  Block rules: {N}

**Policy overrides:** {N} active

Follow with a table of governance rules and overrides if any exist.

---

General rules

• Always read the existing profile before making changes. Never overwrite fields the user did not ask to change.
• Always update .timestamp to the current ISO-8601 time on every write.
• Never set .schema to any value other than 1.
• When adding audit entries, always require a reason. Do not allow empty reason strings.
• If the user tries to set a severity value other than pass, warning, or fail, reject it and list the valid options.
• If the user tries to ignore all six assessments simultaneously, warn them that this would result in every package passing regardless of content.
• After every write, confirm the file path and summarize only the fields that were changed.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: reversinglabs
• Source: reversinglabs/rl-protect-skills
• License: MIT
• Homepage: https://secure.software/

Install and usage instructions live in the source repository linked above.