No description
Find a file
gremlin a0c09c69d2 gsds: add exporter-warn-log-for-intended-auth-fronted-probe-behavior skill
New skill (checked novel, embedded in skills_v1) capturing the
blackbox_exporter/Authentik redirect-noise pattern from this session's
Blackbox audit.

Also fixes a pre-existing YAML syntax bug in the stash findDuplicateScenes
ordering-drift skill (plain scalars containing mid-line ': ' are ambiguous
to PyYAML) that was blocking ingest_skills.py's full-directory scan --
content unchanged, quoting/wrapping only.
2026-08-04 00:42:14 +00:00
command-router docker domain: document new updates/update actions 2026-07-27 22:55:34 +00:00
galds-v1 galds-v1: sync doc-request source split from GALDS-v1-Schema-Spec.md 2026-07-15 21:27:47 +00:00
gbds-v2 reports-v1: document source/diagnosis/diagnosed_at fields for Gremlin Diagnose 2026-07-17 01:03:26 +00:00
ghcs-v1 chore: add synced_from headers to schema skills 2026-07-02 16:37:16 +00:00
gsds gsds: add exporter-warn-log-for-intended-auth-fronted-probe-behavior skill 2026-08-04 00:42:14 +00:00
imp-actions skills: add pocket-adapted skills (imp-actions logging, pocket-stack-standard, pocket-status, pocket-command-bus) 2026-07-05 19:46:34 +00:00
pocket-command-bus docs: consolidate sync docs — drop Sync SSD and family syncoid entry 2026-07-30 19:27:56 +00:00
pocket-stack-standard docs: add 6xxx host-port convention to pocket-stack-standard 2026-07-30 20:53:17 +00:00
pocket-status skills: add pocket-adapted skills (imp-actions logging, pocket-stack-standard, pocket-status, pocket-command-bus) 2026-07-05 19:46:34 +00:00
reports-v1 reports-v1: sync capacity-audit type from Reports-Schema-Spec.md 2026-08-02 01:26:58 +00:00
swarm-stack-standard skill: sync command-router + swarm-stack-standard for gremlin.diagram.* 2026-07-19 00:06:29 +00:00
README.md GSDS: fix category taxonomy -- re-categorize both skills out of the infra catch-all 2026-07-17 22:32:04 +00:00

skills

This repo serves two distinct, non-overlapping purposes:

  • <skill-name>/SKILL.md — Claude Code's own loadable skill library. Symlinked live at ~/.claude/skills on gremlin-worker; the harness reads each SKILL.md's frontmatter (name, description) to build the per-session "available skills" list. Every top-level directory except gsds/ follows this pattern.
  • gsds/**/*.yaml — GSDS (Gremlin Skill Data Schema) procedural-memory records: "how Gremlin solved class-X problems before," retrieved via semantic search (Qdrant skills_v1), not loaded by the Claude Code harness. Structurally inert to the SKILL.md mechanism above (different filename, no frontmatter the harness looks for) — confirmed empty by design, not by accident. See traveler/memory/2026/07/2026-07-17-gsds-skills-collision-resolution.md for how this split was decided.

Don't add a new top-level directory for either system without checking which convention it should follow.

GSDS category taxonomy

gsds/<category>/<skill_id>.yaml — the category enum (also the category field value, always matching the directory a skill actually lives in):

infra, docker, n8n, qdrant, wikijs, forgejo, caddy, galds, media, stash, pocket, security, cicd

infra is for genuinely host/hardware/OS-level concerns (ZFS, memory, disk, host-level networking) — it is NOT a catch-all for "anything vaguely infrastructure-related." Docker Swarm orchestration specifics belong in docker; n8n workflow/sandbox specifics belong in n8n; and so on down the list. Both current skills were originally miscategorized as infra and corrected to docker/ n8n on 2026-07-17 once this distinction was made explicit — see traveler/memory/2026/07/2026-07-17-gsds-phase2-build.md.