Skip to content

charlieviettq/awesome-agent-skill / .claude/skills/autoskill

autoskill

Verified

Observe the user's screen via screenpipe, detect repeated research workflows, match them against existing scientific-agent-skills, and draft new skills (or composition recipes that chain existing ones) for the patterns not yet covered. Use when the user asks to analyze their recent work and propose skills based on what they actually do. Requires the screenpipe daemon (https://github.com/screenpipe/screenpipe) running locally on port 3030 — the skill has no other data source and will refuse to run if screenpipe is unreachable. All detection runs locally; only redacted cluster summaries reach the LLM.

config.yaml

View on GitHub
# autoskill configuration
#
# LLM backend for skill synthesis. Detection/clustering always runs locally;
# only redacted cluster summaries are sent to the LLM.
#
# Local is the default — your screen content never leaves the machine.
# Cloud backends (claude, foundry) are available for users who explicitly
# opt in; see the backend sections below.
backend: local    # local | claude | foundry

# Per-backend settings. Only the selected backend's block is used.
local:
  # LM Studio exposes an OpenAI-compatible server. Start it from the
  # "Developer" tab; the default port is 1234.
  endpoint: http://localhost:1234/v1
  # Gemma-4-31B-it is the recommended default: strong reasoning at a size
  # most modern workstation GPUs can run. Swap for any LM Studio model ID.
  model: Gemma-4-31B-it

claude:
  model: claude-opus-4-7
  # api_key read from ANTHROPIC_API_KEY env var

foundry:
  endpoint: https://foundry.example.com/anthropic
  model: claude-opus-4-7
  # api_key read from FOUNDRY_API_KEY env var

# Screenpipe HTTP endpoint. For TLS, point this at your local Caddy proxy
# (see references/https-proxy.md).
screenpipe:
  url: http://localhost:3030
  # Screenpipe requires a bearer token for its local API. Either set `token`
  # here, or export SCREENPIPE_TOKEN in your environment (preferred — keeps
  # the token out of version control). Retrieve with: `screenpipe auth token`.
  # token: your-token-here

# Embedding model for matching against existing scientific skills.
# Local only; no API calls.
embeddings:
  model: sentence-transformers/all-MiniLM-L6-v2

# Clustering thresholds.
cluster:
  min_session_minutes: 5        # skip sessions shorter than this
  idle_gap_minutes: 10          # new session after this much inactivity
  min_cluster_size: 2           # need this many similar sessions before proposing

# Content redaction regexes applied before any cluster summary leaves
# the local detection layer. Defense-in-depth on top of screenpipe's
# own app/window filtering.
redaction:
  enabled: true
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexUnverifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall