Agent skills for coding CLIs, multi-agent runtimes, context engines, MCP extensions, and terminal tooling. Instead of using claude code's source code, give your agent skills to create your own!
Skill Claude CodeCodex
Release every agent-scoped resource in the finally block so spawning, aborting, or finishing an agent leaves no lingering clients, hooks, caches, or background tasks.
Skill Claude CodeCodex
Merge local and MCP skill sources into one listing surface and dedupe by stable command name before announcing them.
Skill Claude CodeCodex
Skip the background memory extractor when the main agent already persisted memory during the same turn to avoid duplicates and race hazards.
Skill Claude CodeCodex
Normalize the main thread to a concrete empty-string key so agent-scoped state maps treat main and forked threads uniformly.
Skill Claude CodeCodex
Gate cached microcompact cache edits to the main thread so background forks cannot alter the shared cache state.
Skill Claude CodeCodex
Filter prefetched memory attachments against cumulative read state before mutating that state, so async surfacing does not dedupe itself away.
Skill Claude CodeCodex
React to marketplace materialization diffs by refreshing plugins when new installs land and escalating to needsRefresh when updates or refresh failures occur.
Skill Claude CodeCodex
Ensure SkillTool only exposes MCP skills when they are registered as prompt-based commands and dedupe them against local commands.
Skill Claude CodeCodex
Sanitize user messages by replacing images, documents, and other heavy media blocks with lightweight placeholders before compaction.
Skill Claude CodeCodex
Load independent command sources in parallel, then memoize the merged registry by cwd so repeated lookups stay cheap and ordered.
Skill Claude CodeCodex
Ensure memory directories exist once and surface consistent guidance so agents skip redundant mkdir calls.
Skill Claude CodeCodex
Keep MEMORY.md terse by capping entries, warning when caps hit, and pointing each memory to its own file.
Skill Claude CodeCodex
Persist binary output with a mime-derived extension and a matching saved-file message so downstream tools keep the right modality.
Skill Claude CodeCodex
Inject mode-specific skip rules into a shared reconciler so unsupported source types are recorded and skipped without forking the core install logic.
Skill Claude CodeCodex
Resolve skill names through layered namespace lookup, then preload the valid skills as visible startup messages.
Skill Claude CodeCodex
Process every nested-memory trigger before wiping the trigger set so subsequent iterations only see new paths.
Skill Claude CodeCodex
Return no attachment when an incremental skill delta has no unseen items instead of emitting a formally empty update payload.
Skill Claude CodeCodex
Map async marketplace reconciliation events into AppState rows so UIs show install/installing/failed badges without blocking.
Skill Claude CodeCodex
Commit the connected state first, then gather optional capabilities and send a post-connect broadcast asynchronously without delaying readiness.
Skill Claude CodeCodex
Validate remote numeric override maps defensively so null, non-finite, or negative flag payloads never corrupt runtime thresholds.
Skill Claude CodeCodex
Buffer and execute streaming tool calls with concurrency awareness so results appear in request order even when multiple runners overlap.
Skill Claude CodeCodex
Remove assistant messages that reference tools without recorded results before feeding the history back into runAgent.
Skill Claude CodeCodex
Design skill steps with explicit execution topology, asking which parts run directly, in subagents, in teammate swarms, or in parallel.
Skill Claude CodeCodex
Keep skill-listing deltas scoped per agent or thread so each agent gets its own first-turn announcement surface.
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: