slicing-long-contexts
Use when a prompt or corpus is long/dense (multi-docs, logs, codebases) and you want a reproducible map/reduce pipeline. Trigger this skill to slice inputs, run per-slice codex/gemini subcalls, and aggregate results with manifests/logs via the slice runner (preferred default), falling back to manual REPL slicing only if needed.
npx skills add majiayu000/claude-skill-registry --skill slicing-long-contexts --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
# RLM CLI Runner Use this skill to replicate the paper's REPL-based RLM pattern: treat the long prompt as data in a Python REPL, peek/slice it with code, and spawn recursive sub-LM calls (codex or gemini) on targeted snippets. Designed for dynamic context (write big outputs to files; read with tail/rg) and AGENTS preferences (plans/logs/results outside the skill dir). Trust posture: ASK for writes/network; keep sandbox workspace-write unless a task requires more. `--with-network` toggles codex/gemini network; leave it off unless needed. ## Quick start (one command) ```text python <CODEX_HOME>/skills/slicing-long-contexts/scripts/slice_runner.py --prompt <corpus-file> --question "<target>" --provider openai --chunk-size 30000 --prefer-headings --max-slices 6 --out-dir rlm_outputs/<run_id> --run-id <run_id> --with-user-codex-access --summary-cmd-template "codex {approval_flags} exec --model {model} \"$(cat {prompt_path})\"" --summary-system-prompt "You are writing <target-doc>. Combine sub-responses into a concise, structured, actionable output." --summary-out rlm_outputs/<run_id>/rlm_summary.txt ``` Replace `<CODEX_HOME>` with your installed skill root (for example, `~/.codex` or `C
- Quick start (one command)
- Decision trigger (use the runner by default)
- Default path (runner, use this first)
- When to use
- Inputs / Outputs
- Trust / bounds
- Prereqs
- Workflow (plan → instrument → execute → verify)
- Advanced / manual REPL workflow (only if runner is insufficient)
- Patterns to reuse
- References
- Default vs advanced usage
- Usage notes
- Common commands
What does the slicing-long-contexts skill do?
Use when a prompt or corpus is long/dense (multi-docs, logs, codebases) and you want a reproducible map/reduce pipeline. Trigger this skill to slice inputs, run per-slice codex/gemini subcalls, and aggregate results with manifests/logs via the slice runner (preferred default), falling back to manual REPL slicing only if needed.
How do I install it?
Run `npx skills add majiayu000/claude-skill-registry --skill slicing-long-contexts --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From majiayu000/claude-skill-registry, a repository with 534 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.