EXTENSIONS

Semantic discovery index for published swamp extensions. Crawls the extension registry, embeds metadata using an OpenAI-compatible API, and provides hybrid keyword + vector search (FTS5 + cosine, RRF-fused) to find extensions by intent rather than exact name.

Local git operations model — clone, branch, read, commit, push. Designed for agent-driven development workflows where code changes are authored locally and pushed to a remote forge. Supports a configurable workspace layout (default: $HOME/{host}/{group}/{project}) and commit message format enforcement.

Fact-check a single documentation file. Its job is to confirm that the FACTS

Scrapes Prometheus metrics from Coder's metrics endpoint. Extracts key operational indicators (API latency, workspace counts, agent connections) and writes typed snapshots for trend analysis.

Queries Coder workspace API and captures state snapshots. Records workspace status, agent state, and build info for stability and drift analysis.

Pages through the Coder audit log API and writes events as versioned data. Supports incremental collection with configurable limits and query filters.

Records webhook payloads received from Coder notifications. Stores each event with receipt metadata for delivery latency and reliability analysis.

Polls Coder server health and build info endpoints. Produces typed health state with latency measurements for reliability analysis.

A daily operator briefing workflow with companion report. Gathers data from GitLab (review queue, todos), Microsoft Teams (mentions, unread chats, channel messages), and Redmine (recent updates, new tickets), then renders a prioritized markdown briefing plus a stable JSON contract.

Automated web reading pipeline — harvests articles from configurable sources (Hacker News, Lobsters, RSS/Atom feeds), evaluates them with an AI agent, learns user interest over time via feedback, and produces a curated HTML report with ranked recommendations. Designed for agent-driven reading workflows where an AI reads on the user's behalf and surfaces what matters.

Fleet-wide Kubernetes health across all kubeconfig contexts. Iterates every cluster in ~/.kube/config (or a specified subset) and produces per-cluster health snapshots plus an aggregated fleet summary. Best-effort — unreachable clusters are reported, not fatal.

One-shot bootstrap for @twonines/fact-store backed by AWS Aurora Postgres Serverless v2. Provisions the RDS cluster + writer instance, the enclosing VPC networking primitives (DB subnet group, security group), an IAM managed policy scoped to rds-db:connect on the cluster, and an IAM workload role that trusts a caller-provided principal (e.g., an AWS SSO permission set ARN). A companion workflow runs the provisioner and then configures @webframp/postgres-datastore against the newly created cluster using a fresh RDS IAM auth token.

Stores, validates, and serves organizational facts for AI agent consumption. Supports a propose→review→activate lifecycle with adversarial validation (ferret/mole pattern). Facts are relational truths about infrastructure, repositories, services, teams, and their connections. Ships with three agent skills — propose-facts and review-proposals drive the discovery and review loops (ferret/mole); consult-facts closes the loop as the consumer role, querying facts before engineering work and detecting drift.

Clones a GitLab repository, chunks all text files, embeds them via an OpenAI-compatible API, and writes a SQLite database supporting hybrid search (FTS5 keyword + vector cosine with RRF reranking). Methods: index, index_batch, search, reindex, status, discover, list-indexed, list-searches.