ericrisco/rsc-harness / skills/unslop
unslop
VerifiedUse when a text that leaves the session (email, post, README, message, landing copy) must sound like a person and carry no AI tells: rewrites it in the user's language fitted to the channel and relationship, and audits it against a named catalogue of tells (puffery, `not just X but Y`, rule of three, em dashes, filler, hedging). Fires on 'bro', 'hazlo más humano', 'que no suene a IA', 'más natural', 'unslop this'. NOT how the harness talks to the user (that is `orient`), NOT a reusable voice guide (that is `brand-voice`).
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 9.3 KB |
| evals/README.md | markdown | 1.3 KB |
| evals/cases.yaml | config | 3.8 KB |
evals/README.md
View on GitHub# Evals — unslop
The claim under test is the boundary as much as the capability. `unslop` absorbed the retired `bro`,
so it owns both the register pass and the audit of text that leaves the session. Its neighbour is
`orient`, which owns how the harness talks to the user. The negatives prove that split holds.
## Triggering
Load only `unslop` and run every prompt in 3–5 fresh sessions. A positive passes when `unslop` loads
in a majority of trials; a negative passes when it stays quiet and the declared `route_to` skill is
the real owner. Require at least 90% accuracy, and treat any failure on the `orient` negative or on
the two former `bro` positives as blocking rather than statistical.
Run the mirror direction too. `orient`'s own suite carries a negative that routes here; both must
pass in the same round, or the boundary only exists on one side.
## Capability
Run the scenario without the skill and with it, same post and one grader, three trials. The
with-skill condition must satisfy at least 80% of the rubric and beat the baseline by at least 20
percentage points.
Two points discriminate hardest: **naming** the tells by category (a baseline rewrites silently) and
**preserving** the code block, the price and the URLs (a baseline edits them while tidying). Grade
both strictly.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |