From temper
Detects project stack and enforces quality gates (SUGGEST/WARN/BLOCK) with confidence scoring, adaptive learning, review memory, and capabilities like architecture depth and grill me.
How this skill is triggered — by the user, by Claude, or both
Slash command
/temper:temper-coreThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Stack detection → Quality gates (SUGGEST/WARN/BLOCK) → Confidence scoring (0.0-1.0) → Review memory → Adaptive learning → Metrics.
Stack detection → Quality gates (SUGGEST/WARN/BLOCK) → Confidence scoring (0.0-1.0) → Review memory → Adaptive learning → Metrics.
.claude/temper.config → stack field.claude/presets/*.yaml → stack section.claude/packs/stacks/{stack}.mdThree-tier: project-local > global > built-in. Cached in .temper/pack-manifest.json and consumed by all stage commands (build, review, check, plan, design) for phase-filtered loading.
.claude/packs/{name}/rules.md (project)~/.claude/packs/{name}/rules.md (global)$CLAUDE_PLUGIN_ROOT/.claude/packs/{name}/rules.md (built-in)
Packs support link: plugin://name | skill://name and phases: [build, review, ...]..temper/review-memory.json — auto-suppress after 5 dismissals.temper/metrics.jsonPost-review intelligence layer that makes reviews smarter over time. Runs as Step 8.5 in /temper:review and displays in /temper:status.
Three capabilities:
| Capability | Trigger | Action |
|---|---|---|
| Pattern Detection | Post-review hook (Step 8.5) | Cluster recurring findings by category + file pattern + keywords |
| Rule Suggestions | Pattern accepted 3+ times at 70%+ rate | Generate pack rule template, queue for promotion |
| Noise Reduction | Pattern dismissed 5+ times | Auto-suppress, reduce false positives |
Key files:
.temper/learning.json — Learning state (detected patterns, suppressions, suggestions, learning curve).temper/learning/suggestions/ — Generated rule suggestion templates.claude/packs/adaptive-learning/rules.md — Promoted rules (user-accepted suggestions)Graceful degradation: If learning.json is absent, all commands work unchanged. No errors, no warnings.
Full docs: $CLAUDE_PLUGIN_ROOT/.claude-plugin/reference/learning.md
Four independently optional capabilities, all enabled by default. Controlled via capabilities section in temper.config.
| Capability | Stage | Purpose |
|---|---|---|
| Architecture Depth | Review | Module-depth analysis: seams, adapters, locality, leverage, deletion test |
| Grill Me | Plan, Design | Socratic challenge mode — stress-test plans before building |
| Config Suggestions | Check | Suggest CLAUDE.md/AGENTS.md updates based on what was built |
| HTML Review | Plan | Interactive browser-based plan review with inline comments |
Graceful degradation: Each capability checks its config flag. Missing config = all enabled (default-on).
$CLAUDE_PLUGIN_ROOT/.claude-plugin/reference/{command}.md
npx claudepluginhub galando/temper --plugin temperConducts tiered code reviews for security (OWASP top 10), performance, and quality on staged git changes, PRs, or files before commits or releases.
Assesses codebase for AI agent readiness by detecting stacks, monorepos, git setup, and evaluating style, testing, code quality, secrets, and file sizes.
Assesses codebases against Arness patterns in CLAUDE.md, prioritizes improvements via decision gates, and orchestrates multi-spec pipeline from planning and execution to testing and shipping.