[RAGE-8][P1] Discover and import external skills into the canonical native Prolog skill path #40

New issue

Open

opened 2026-09-20 01:40:01 +00:00 by nsaspy · 0 comments

nsaspy commented

2026-09-20 01:40:01 +00:00

Owner

Parent: #6
Consumes: lost-rob0t/prolog-rlm#173
Feeds: RAGE-6 / RAGE-7

Goal

Discover supported external skill packages and normalize them into the canonical native representation before evaluation/activation.

Initial sources

standard Agent Skills / SKILL.md
Claude / Claude Code
OpenCode
Agent Zero
AgentProlog-native skill directories

Requirements

preserve source provenance and original package identity;
lenient source parsing only where the upstream #173 contract permits it;
normalize strictly into the canonical internal representation;
missing routing/graph semantics use documented deterministic defaults or later trusted enrichment;
imported requests for tools/capabilities/authority do not grant anything;
imports enter the same evaluation/lifecycle path as generated candidates;
avoid multiple independently authoritative copies after import;
reverse export is explicitly out of scope.

Acceptance

at least standard + two compatibility variants import through the upstream contract;
provenance remains inspectable after normalization;
imported skill can be rejected without entering active state;
duplicate package/version imports resolve deterministically;
no external format constrains richer native Prolog semantics.

Parent: #6 Consumes: lost-rob0t/prolog-rlm#173 Feeds: RAGE-6 / RAGE-7 ## Goal Discover supported external skill packages and normalize them into the canonical native representation before evaluation/activation. ## Initial sources ```text standard Agent Skills / SKILL.md Claude / Claude Code OpenCode Agent Zero AgentProlog-native skill directories ``` ## Requirements - preserve source provenance and original package identity; - lenient source parsing only where the upstream #173 contract permits it; - normalize strictly into the canonical internal representation; - missing routing/graph semantics use documented deterministic defaults or later trusted enrichment; - imported requests for tools/capabilities/authority do not grant anything; - imports enter the same evaluation/lifecycle path as generated candidates; - avoid multiple independently authoritative copies after import; - reverse export is explicitly out of scope. ## Acceptance - [ ] at least standard + two compatibility variants import through the upstream contract; - [ ] provenance remains inspectable after normalization; - [ ] imported skill can be rejected without entering active state; - [ ] duplicate package/version imports resolve deterministically; - [ ] no external format constrains richer native Prolog semantics.

nsaspy referenced this issue

2026-09-20 01:40:38 +00:00

[EPIC] Personal skill refinery: observe workflows, synthesize candidates, RAGE-evaluate, and import external skills #6