Vahlame
UserCross-platform kit that gives AI agents (Claude Code, Codex, Cursor) persistent memory: Markdown vault + git + MCP. Local-first, hybrid search optional.
Categories
Indexed Skills (7)
vkm-design
Professional, anti-generic design for REAL design work — a new surface, a redesign, an identity, dataviz, or a design critique. Forces a 3-line brief, a named direction, computed checks and verified-current libraries. NOT for one-line style tweaks, or logic with no visible surface.
vkm-discipline
Use when implementing, fixing, refactoring, migrating, debugging, deploying, cleaning data, reviewing a PR, or writing a postmortem/README — reads intent, scales depth to stakes, ends in executed evidence. NOT one-line single-file edits (typo, bump), questions, explainers, or scope the user set.
vkm-research
Consolidates a RESEARCH/<topic> bank into a quality summary.md — reads hub + unconsolidated sources, synthesizes with wikilinks + supersedes, validates, marks done. Imports external reports. Invoke for "consolidate <topic>", "write up the research on X", or after obscura_research persisted sources.
vkm-spec
Turns a vague idea into a precise, testable, vault-grounded spec before any code — one assemble_context call, a filled template, machine validation, human review. Use when the user says "spec this", "write a spec", "plan this feature", or hands over an idea that needs requirements first.
vkm-intake
Task intake BEFORE non-trivial execution. Restate objective, deliverable and non-goals in 3 lines, one closed question max on ambiguity, inventory what attached images actually show, assemble minimal context, then execute exactly that scope. NOT for one-line edits or tasks already scoped.
vkm-seo
Use when building, auditing, or optimizing how a website ranks in search and AI answers — "audit my SEO", "why don't we rank", meta/title/schema fixes, local/multi-location pages, synonym/intent coverage, GEO/AEO. Ships a before/after audit. NOT paid ads/SEM, social, rank tracking, or non-web.
vkm-verify
Prove a green check actually ran, covered the change, and can fail. Use before claiming done / tests pass / CI green / committed / deployed, when a guard or lint has never once fired, or when a pass contradicts a symptom the user still sees. NOT for writing tests or debugging a known-red failure.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.