{% extends "base.html" %} {% block title %}prune {{ project.project_key }} · watchmen{% endblock %} {% block content %}
← {{ project.project_key }}

Prune skills

LLM-judge flagged skills for review. Approve removes the skill; Dismiss keeps it (and remembers your decision for next time).

{% if queue is none %}
No prune queue yet
Run watchmen prune {{ project.project_key }} to generate one. The judge reads your bundled SKILL.md files, the workspace brief, and per-skill usage telemetry from corpus.db, then flags low-value skills here.
{% elif flagged|length == 0 %}
Queue is empty
Nothing flagged in the last prune run. {% if queue.summary %}
{{ queue.summary }}{% endif %}
{% else %} {% if queue.summary %}
Judge summary
{{ queue.summary }}
{% endif %}
{% for entry in flagged %}
{% if entry.severity == "high" %} high {% elif entry.severity == "medium" %} medium {% else %} low {% endif %} {{ entry.slug }}
{% if entry.description %}
{{ entry.description }}
{% endif %}
{{ entry.reason }}
{% endfor %}
{% endif %} {% if dismissed %}
Previously dismissed ({{ dismissed|length }})
These slugs were kept on a prior prune run. Future judges will be told and can re-flag with new evidence (e.g., still never fired six weeks later) — they aren't permanently suppressed.
{% for slug in dismissed %} {{ slug }} {% endfor %}
{% endif %} {% endblock %}