Skip to content

Feat: Introduce 'Skill Health Score' and Automated Degradation Alerts for Enterprise Governance #587

Description

@FenjuFu

Motivation

Today's GitHub Trending highlights a critical shift in AI engineering: moving from "single-point capability breakthroughs" to "infrastructure standardization and governance". Projects like Canner/WrenAI (16k+ stars) are gaining traction specifically because they address the "Trust and Governance" gap in GenBI—ensuring data lineage and query reliability. Similarly, bojieli/ai-agent-book emphasizes systematic, structured knowledge over fragmented tools.

While iflytek/skillhub currently excels at publishing and RBAC (as seen in existing issues), it lacks a mechanism for post-deployment quality assurance. In enterprise environments, a "published" skill is not necessarily a "healthy" skill. Without automated health monitoring, teams face the "Black Box" anxiety described in today's trend analysis: they cannot distinguish between a broken skill and a slow one, leading to compliance risks and operational instability.

Proposed feature

Introduce a "Skill Health Score" (SHS) system with automated degradation alerts:

  1. Metric Aggregation: Automatically collect runtime metrics for installed skills (e.g., success rate, average latency, error codes) from astron-agent or astron-rpa execution logs.
  2. Dynamic Scoring: Calculate a composite health score (0-100) based on:
    • Reliability: % of successful executions over the last N runs.
    • Performance: P95 latency compared to historical baselines.
    • Stability: Frequency of unexpected crashes or timeout errors.
  3. Governance UI: Add a "Health" tab to the Skill Hub dashboard (similar to WrenAI's semantic layer governance view) showing:
    • Real-time health status badges (Green/Yellow/Red).
    • Trend charts for latency/error rates.
    • Automated Alerts: Notify namespace admins via Webhook/Email when a skill's score drops below a configurable threshold (e.g., < 80).
  4. Deprecation Workflow: Allow admins to "Soft Disable" low-health skills while maintaining audit trails, preventing downstream agents from calling unstable tools.

Why it fits this repo

iflytek/skillhub is positioned as the "Enterprise Skill Registry". Today's trend analysis explicitly states that developers want a "Trustworthy Skill Registry" that offers "Governance" and "Audit Trails".

Existing features handle who can publish (RBAC) and what is published (Versioning). This feature addresses how well it performs. It directly complements the existing Skill Provenance and SAST features by adding a runtime observability layer. This transforms SkillHub from a passive repository into an active governance platform, aligning with the "Infrastructure Standardization" theme of today's trending repos.

References

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    effort/s小改动或边界明确 / Small or well-bounded change.priority/p2中优先级 / Medium priority triage bucket.triage/needs-info需要补充更多信息后才能分流 / Issue needs more detail before it can be routed.

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions