Skill Upgrade Checker

Primary role: Scan all installed skills for available upgrades, interpret version differences and changelogs, assess upgrade risk, and execute upgrades after explicit user confirmation.

主要职责： 扫描所有已安装技能的可用升级，解读版本差异与变更日志，评估升级风险，经用户明确确认后执行升级。

Prerequisites

Before this skill can operate, confirm the following:

• - clawhub CLI is installed and available in the agent's PATH. This skill issues clawhub list, clawhub search, and clawhub update commands. If the binary is absent, all operations will fail.
• Authentication: Run clawhub login in advance. The clawhub CLI stores credentials in its own default config path (managed by the CLI, not by this skill). No additional environment variables are required by this skill.
• Permissions in scope: This skill will only:

1. Read the installed skill list via clawhub list (read-only). 2. Search the registry for latest versions via clawhub search <skill> (read-only). 3. Run clawhub update <skill> — only after explicit user confirmation at the Pre-Upgrade Gate (Step 6). It will not access other system files, credentials, or APIs beyond the clawhub CLI.

Workflow

0) Inventory Installed Skills

Run clawhub list and parse the output to build a table of all installed skills with their local version numbers.

Additionally, read each skill's _meta.json (at skills/<skill-slug>/_meta.json) if accessible locally, to get the precise semver from the version field.

Output: a list of installed skills with their local versions.

1) Query Registry for Latest Versions

For each installed skill, run:
CODEBLOCK0

Build a comparison table with columns:

• - Skill name
• Local version
• Registry version
• Upgrade available (yes / no / unknown)

If clawhub search does not return version info for a particular skill (e.g., it is a local-only skill not published to any registry), mark it as "version unknown — skipped".

2) Version Comparison and Classification

Apply semver comparison to each skill where an upgrade is available. Classify into:

• - Patch (e.g., 0.2.1 → 0.2.2): bug fixes only, low risk
• Minor (e.g., 0.2.x → 0.3.0): new features, backward-compatible, medium risk
• Major (e.g., 0.x → 1.0 or 1.x → 2.0): potential breaking changes, high risk

Edge cases:

• - Local version higher than registry → mark as "local version ahead" and skip
• Versions equal → mark as "up to date"

Present a summary table sorted by risk level (major first).

3) Changelog and Diff Interpretation

For each skill with an available upgrade, gather change information in this priority order:

1. 1. Project-level CHANGELOG: Check skills/CHANGELOG.md for entries between the current and latest version.
2. Per-skill CHANGELOG: Check skills/<skill-slug>/CHANGELOG.md if it exists (more specific).
3. Description diff: Compare the description field in local _meta.json with the description returned by clawhub search. Changes in description may hint at capability changes.
4. Fallback: If no changelog data is available from any source, state: "No changelog available — manual review recommended before upgrade."

Summarize changes per skill:

• - Added: new capabilities
• Changed: behavior modifications
• Fixed: bug fixes
• Breaking: incompatible changes (flag prominently)

4) Risk Assessment

For each upgradeable skill, produce a risk rating using these criteria:

Dimension	LOW	MEDIUM	HIGH	CRITICAL
Version jump	Patch only	Minor	Major	Major with breaking keywords
Breaking change signals

None found | "changed", "refactored" | "deprecated" | "breaking", "removed", "incompatible" |
| Dependency impact | No other skill depends on it | 1 skill depends | 2+ skills depend | Core dependency for many skills |
| Version staleness | < 1 month behind | 1–3 months | 3–6 months | > 6 months behind |
| Adoption signals | High install count | Moderate | Low (< 10) | Very new (< 2 weeks) or unknown |

Combine dimensions into an overall rating: LOW / MEDIUM / HIGH / CRITICAL.

5) Upgrade Recommendations

Present a prioritized recommendation list, grouped into four tiers:

• - Recommended to upgrade now (推荐立即升级): low-risk patches and well-established minor updates
• Upgrade with caution (谨慎升级): minor updates with some risk signals
• Review before upgrade (升级前需审查): major updates or high-risk changes
• Skip (跳过): not worth upgrading, risk outweighs benefit, or local version ahead

Include a one-line rationale per skill.

6) Pre-Upgrade Confirmation Gate (required)

Before executing any upgrade, present a confirmation summary listing:

• - Each skill to be upgraded
• From-version → to-version
• Upgrade type (patch / minor / major)
• Risk rating
• One-line change summary
• The exact clawhub update <skill> command to be run

Then ask:

"Planned action: upgrade N skill(s). Confirm which to proceed with — respond with skill names, 'all', or 'none'."

Major upgrade extra gate: Even if the user responds with "all", major-version upgrades require individual confirmation per skill. Re-prompt for each major upgrade separately with a clear warning about potential breaking changes.

7) Execute Upgrades and Post-Upgrade Report

For each confirmed skill, in sequence:

1. 1. Run INLINECODE20
2. Verify success by re-running clawhub list and checking the new version
3. If any upgrade fails, stop and report the failure — do not continue blindly

Final output:

Upgrade Results Table:

Skill	Old Version	New Version	Status
...	...	...	Success / Failed

Post-Upgrade Summary:

• - Updated skill inventory (full list with current versions)
• Any failures with error details and suggested next steps
• Follow-up suggestion: "Re-run this check periodically to stay current."

Output Format

Upgrade Overview Table

| Skill | Local Version | Latest Version | Upgrade Type | Risk Level | |-------|---------------|----------------|--------------|------------|

Change Summary (per skill)

• - Added: ...
• Changed: ...
• Fixed: ...
• Breaking: ...

Recommendation List

Grouped by tier (Recommended / Caution / Review / Skip) with one-line rationale.

Pre-Upgrade Confirmation

"Planned action: upgrade skill(s). Source: clawhub. Risks:

. Proceed with: ?"

Upgrade Results

| Skill | Old Version | New Version | Status |

Guardrails

• - Never upgrade without explicit user confirmation.
• Major-version upgrades require individual confirmation even if user says "all".
• If an upgrade fails, stop and report — do not continue blindly.
• Audit first, act second — Steps 0–5 are entirely read-only.
• If local version is ahead of registry, skip and note it — never downgrade.
• If no changelog is available, recommend manual review before proceeding.
• Prefer conservative risk ratings — when uncertain, rate higher rather than lower.