Releases: handlecusion/nah-mean
Release list
v0.5.0
Full Changelog: v0.4.0...v0.5.0
v0.4.0
Adds Korean post-work feedback triggers for intent recovery and reinforcement.
- Added
감다뒤recovery: after a disliked result, agents restate the original intent, identify the missed sight, and align a corrected standard before rework. - Added
감다살reinforcement: after accurate intent capture, agents restate the matched intent and reinforce the standard in runtime memory. - Updated README, prompts, adapters, AI discovery files, docs, security install guidance, and pinned install examples to
v0.4.0.
v0.3.2
Cleans repository packaging by removing deterministic benchmark scripts, baselines, and generated result files. README and AI discovery now use a simpler tradeoff note instead of numeric benchmark claims.
v0.3.1
Corrects the benchmark baseline to vanilla agent behavior with no skill installed. Results now state that vanilla wins upfront token overhead, while nah-mean improves expected rework units by 81.8%, coordination workload units by 38.2%, and estimated rework-adjusted seconds by 49.8% on the included deterministic fixture set.
v0.3.0
Adds light-align + context-fit execution dispatch. Korean docs now position the skill as 뭔말알-알잘딱-Driven. Includes deterministic benchmark fixtures showing a 50.0% reduction in alignment workload units and 34.3% lower estimated review time versus v0.2.1 baseline.
nah-mean v0.2.1
Documentation split release. README.md is now English-only for English-speaking users, README.ko.md contains the Korean documentation and Korean-specific use cases, and AI discovery metadata/docs links were updated to reference both language versions.
nah-mean v0.2.0
GEO readiness release. Adds MIT license, security and contribution docs, citation metadata, changelog, llms.txt, llms-full.txt, enriched manifest metadata, README answer blocks and FAQ, GitHub Pages-ready docs with JSON-LD, and packaged skill release asset with SHA256 checksum.
nah-mean v0.1.0
Initial public release of the nah-mean intent alignment skill. Includes Codex skill package, bilingual prompts, generic adapters, and install instructions.