Skill Workshop — Coding Agent Skills Review and Installation
Skill Workshop — Overlap Analysis
Eight proposed coding agent skills were reviewed against existing workspace capabilities. The key finding: git-workflow overlaps with gatekeeper. The gate hook mechanism in git-workflow would conflict with gatekeeper's existing static validation, atomic rollback, and git checkpointing pipeline. Recommendation: extend gatekeeper with parity-check triggers instead of creating a parallel pre-commit system.
Approved for Install (7 skills — no overlap)
| # | Skill | Justification |
|---|---|---|
| 1 | pylint/flake8 | No existing PEP-8/type annotation lint capability |
| 2 | html-validate | No HTML well-formedness validator in current toolchain |
| 3 | makefile-generator | Substrate-gate is a specific pipeline; doesn't manage build targets generically |
| 4 | ssh-orchestration | Ansible covers provisioning; no ad-hoc SSH/fabric/invoke task runner |
| 5 | pytest-writer | Eval-harness-runner executes/scores; complementary (one writes scaffolding) |
| 6 | config-management | Reconcile-manifest detects drift but doesn't manage .env/config files |
| 7 | deploy-verify | No post-deploy HTTP health check or built-vs-deployed HTML parity validation |
Skip/Reconsider (1 skill)
git-workflow: Partial overlap with gatekeeper. Gate hook mechanism conflicts with gatekeeper's existing static validation + atomic rollback + git checkpointing pipeline. Recommendation: extend gatekeeper with parity-check triggers instead of creating a parallel pre-commit system.
Existing Skills Referenced During Analysis
- gatekeeper (skills/gatekeeper) — safe mutation staging, static syntax validation, atomic rollbacks, two-pass idempotency, git checkpointing
- substrate-gate (skills/substrate-gate) — pipeline orchestrator (kt-boundary to model inference to ADR to gatekeeper canary)
- eval-harness-runner (skills/eval-harness-runner) — structured evaluation execution and scoring
- reconcile-manifest (skills/reconcile-manifest) — agent state vs manifest drift detection
- workspace-steward (skills/workspace-steward) — token limits, KPI metrics, context pruning
Workshop Proposal Lifecycle Summary
The Workshop proposal lifecycle creates pending SKILL.md drafts that require explicit approval before activation. This session's activity:
| Status | Skill | ID Suffix | Action |
|---|---|---|---|
| Acknowledged | pylint-flake8 | efaa23a1ff | Applied via skill_workshop tool |
| Acknowledged | html-validate | 13b9659cb7 | Applied via skill_workshop tool |
| Acknowledged | pytest-writer | 00fc30e3e5 | Re-created (expired approval) |
| Acknowledged | config-management | deb16dda74 | Re-created (expired approval) |
Blowfish Survey Results (OpenBSD 7.8 amd64)
Ground truth collected during the ansible-role-testing skill verification:
- User: franklin, uid 9001, groups: wheel/operator/www/minecraft
- Services: No rcctl-managed services enabled; /etc/rc.conf defaults to all NO
- MySQL: Running as _mysql:502, PID 76369, ~138MB RSS (pkg installed at /usr/local/)
- X11 workstation: Xorg + xenodm + Openbox, session since September 4
- Pkg forest: Heavy with multiple autoconf/automake versions and GNOME/adwaita deps
- Auth: SSH key auth rejected; password auth works with provided credentials
Ansible-Role-Testing Skill Decision
The ansible-role-testing skill proposal was confirmed as an update to existing content, not a new creation. Ground truth verified via SSH to stargate confirms:
- Host: stargate.research.bitsmasher.net (10.10.16.66)
- User: franklin, password fallback via sshpass
- ansible-test version: 2.19.2 confirmed available
- Collection layout: ~/workspace/lab-franklin/ansible/collections/ansible_collections/lab/franklin/
openbsd-manager Skill Decision
The generic openbsd-manager proposal was declined because blowfish is a GUI workstation (pkg-installed MySQL, not daemon server) with no rcctl services to manage. Instead, two focused skills were proposed:
- openbsd-pkg-audit — generic OpenBSD package health monitoring
- mysql-host-health — cross-platform MySQL instance monitoring
Staging Proposals Pending Review
Additional proposals exist in staging/proposals/ that need review:
| Skill | Status |
|---|---|
| openclaw-config-validation | Create — validate openclaw.json before applying agent edits |
| gradebook-json-validation | Create — gradebook JSON validation for ISMG-2050 |
| documentation-integrity-audit | Create — docs integrity audit |