ericrisco/rsc-harness / skills/postgresdb
postgresdb
VerifiedUse when PostgreSQL engine behaviour decides the answer — schema and type design, index choice, reading EXPLAIN on a slow query, zero-downtime DDL and backfills, or ops (roles, RLS, pooling, vacuum, partitioning, PITR). PG16, ORM-agnostic. NOT portable query logic (that is `sql`), NOT a managed provider's platform surface (that is `neon`).
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 17.3 KB |
| evals/README.md | markdown | 2.6 KB |
| evals/cases.yaml | config | 5.9 KB |
| references/migrations.md | markdown | 11.5 KB |
| references/operations-and-security.md | markdown | 11.5 KB |
| references/query-optimization.md | markdown | 13.2 KB |
| references/schema-and-indexing.md | markdown | 14.1 KB |
| scripts/verify.shexecutable | script | 7.1 KB |
evals/README.md
View on GitHub# Eval harness — `postgresdb`
This harness checks two things: **triggering** (does the skill fire on the right
prompts and stay quiet on near-misses?) and **capability** (does loading the skill
measurably improve the answer?). Cases live in `cases.yaml`. These run through an
**agent harness**, not a pure shell script — a human or a driver agent feeds prompts
and grades outputs. Keep grading honest.
## 1. Triggering
For each prompt in `cases.yaml`:
1. Start a fresh agent session with **only the `postgresdb` skill** discoverable
(its `SKILL.md` description in scope; no sibling skills loaded).
2. Paste the prompt verbatim. Observe whether the agent invokes / routes to
`postgresdb`.
3. Run **3–5 trials per prompt** (the trigger decision is stochastic).
4. Score:
- `should_trigger`: PASS if the skill fires in the majority of trials.
- `should_not_trigger`: PASS if the skill does **not** fire. When the case names
a `route_to` sibling, sanity-check the agent would plausibly route there (or to
`none`); a wrong-but-not-postgresdb route still passes this skill's gate.
**Pass bar:** ≥ 90% trigger accuracy across all `should_trigger` + `should_not_trigger`
prompts (i.e. at most ~1 miss out of the 14 cases). A `should_trigger` that fires < 50%
of trials is a fail; a `should_not_trigger` that fires at all in a majority of trials is
a fail.
## 2. Capability
For each `capability` scenario, run it **twice**:
- **WITHOUT** the skill: a clean agent, no `postgresdb` skill available.
- **WITH** the skill: same prompt, `postgresdb` loaded.
Grade each transcript against that scenario's `must_include` checklist — one point per
bullet that is correctly and specifically covered (not just name-dropped). Compute
coverage = points / total bullets.
**Pass bar:**
- WITH the skill: ≥ 80% of `must_include` bullets covered.
- The skill must **measurably improve** the output: WITH coverage should clearly beat
WITHOUT (target ≥ +30 percentage points, or crossing from fail→pass). If a baseline
agent already nails the rubric without the skill, the rubric is too easy — tighten it.
## Notes
- Prompts are deliberately varied: some never say "Postgres", "SQL", or "index"
(slow-query, multi-tenant isolation, duplicate-job-processing) — these test intent
recognition, not keyword matching.
- Near-misses route to the genuinely correct sibling (`fastapi`, `flutter`,
`deployment`, `security-audit`) or `none` when the request is out of every sibling's
scope (Prisma ORM trap, Redis caching).
- This is judgment-based grading. Record trial counts and the per-bullet verdicts so a
reviewer can audit; don't report a bare pass/fail.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| deprecation-and-migrationaddyosmani/agent-skillsManages deprecation and migration. Use when removing old systems, APIs, or features. Use when migrating users from one implementation to another. Use when migrating a database schema in production, such as renaming or dropping a column without downtime (expand/contract). Use when deciding whether to maintain or sunset existing code. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| performance-optimizationaddyosmani/agent-skillsOptimizes application performance across frontend, backend, queries, and databases. Use when performance requirements exist, when you suspect performance regressions, when Core Web Vitals or load times need improvement, when N+1 query patterns need fixing, or when profiling reveals bottlenecks. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |
| agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |