Rolling context compression - old messages get summarized, recent messages stay verbatim. Never hit the context wall. Toggle it off per session with /rolling-context:off (or --global) when a session wants the full history - subagents follow the session that spawned them, and nothing already compressed is thrown away.
★not rated 31
changed todayA
tokens not measured
originalMIT
v9.29.2 — Model defaults refreshed: Opus 4.7 for planning/strategy/security-review, GPT-5.4 for code-review/implementation. New GPT-5.4 prompting guide. Set OCTOPUSLEGACYROLES=1 to opt out. Run /octo:setup.
★not rated 31▲
+1 3mo agoA
tokens not measured
originalMIT
Skills for AI-native development. /assess scores a codebase's readiness for AI agent contributors (0-8 layered contract model) and generates a Codecov-style complexity hotspot SVG plus a colour-blind-safe doc-navigability graph. /huddle runs structured multi-perspective deliberation using Six Thinking Hats with…
★not rated 30 todayA
tokens not measured
originalApache-2.0
Lists recent Claude Code sessions across every project with summaries, git branch, model, turn count, and a ready-to-paste resume command, and can re-open a whole working set in terminal tabs.
★not rated 30 27d agoA
tokens not measured
originalMIT
Open-source intelligence tradecraft for agents: 29 investigation workflows covering people, companies, domains, images, phones, emails, usernames, crypto and vessels. Every workflow sets an authorised scope before collecting, sources each claim independently, and grades identity attribution separately from the claim…
★not rated 31 1mo agoA
tokens not measured
originalMIT
Two frontier models, one job each. Claude plans and reviews; GPT-5.6 Sol writes the code via Codex CLI. The model that wrote the diff never grades it.
★not rated 27▲
+1
changed yesterdayA
tokens not measured
originalMIT
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: