uditgoenka/autoresearch v2.0.04

v2.0.04 — Reliable Skill Triggering & Probe Recognition

on GitHub

What's New

Reliable Skill Triggering

The autoresearch skill description across all 5 distributions (Claude Code, Claude Plugin, OpenCode, Agents, Codex) has been changed from passive "Use when user types..." to imperative "ALWAYS activate...MUST...BLOCKING". This matches the pattern used by reliably-triggering skills like cook and fix, ensuring autoresearch subcommands activate consistently without requiring users to add emphatic language like "strictly trigger" or "must must trigger."

Root cause: With 150+ skills competing for the model's attention, the passive description allowed Claude to skip skill invocation in favor of handling the request directly. The imperative phrasing makes activation a blocking requirement.

Probe Subcommand Recognition (Codex)

Restored autoresearch-command-spec.json to the Codex skill's resources/ directory. The previous release accidentally deleted this file during a consolidation, breaking the SKILL.md Router's primary lookup path and causing all subcommands (especially probe) to fail recognition when running from the repo source tree.

Codex CLI Hardening

The wrapper CLI (autoresearch_cli.py) now reliably parses all invocation forms:

• $autoresearch:probe — dollar-colon prefix
• $autoresearch probe — dollar-space prefix
• /autoresearch:probe — Claude slash-command compatibility
• autoresearch probe — plain space-separated
• Embedded in prose: "Also run $autoresearch reason --iterations 20"

Full Changelog

• fix(codex): make autoresearch invocations reliable — expanded CLI parser with shlex-based token splitting, prose extraction, and multi-form alias normalization
• fix(codex): restore command spec to skill resources dir — re-added the v2.0.03 spec JSON that was deleted during consolidation
• fix: use imperative skill description for reliable triggering — changed all SKILL.md frontmatter descriptions across 5 distributions
• chore: bump version to 2.0.04

Upgrade

# Claude Code (recommended)
npx @anthropic-ai/claude-code skills install github:uditgoenka/autoresearch

# Manual
./scripts/install.sh --claude --global --force
./scripts/install.sh --codex --global --force
./scripts/install.sh --opencode --global --force