Auto-invoked by sindri-loop on clean termination with ≥1 kept commit. Generates the PR body, pushes the branch, opens a PR via gh, surfaces the URL, and archives the current run. A manual /sindri:finalize (advanced/debug) can force a PR from current kept commits; otherwise do not invoke directly.
EnglishThe sindri orchestrator. Runs one experiment per invocation. Reads state, checks the HALT sentinel and termination predicates, picks next candidate, dispatches an experiment subagent via the Task tool, acts on the result (commit or reset), records to state, and schedules the next wakeup — unless terminated. Invoked automatically via ScheduleWakeup during an active run. A single manual tick is available via /sindri:loop (advanced/debug); otherwise do not invoke directly.
EnglishScaffold .claude/scripts/sindri/benchmark.py by scanning the repo, proposing a measurement in natural language, translating the user's refinements into Python, and running the script once to verify it emits a valid METRIC line. Never ask the user for Python or shell syntax.
EnglishInteractive, one-shot scaffolding for a new sindri optimization run. Parses goal, ensures benchmark.py exists, scans repo for candidates, gets pool approval, calls sindri init (which creates the branch, runs baseline, seeds state), and schedules the orchestrator. Use when the /sindri:forge command receives a new goal statement.
English
AGENT SKILLS ATLASIncludes Sindri Finalize, Sindri Loop, Sindri Scaffold Benchmark, and other workflows.
Claude Code
Codex
Cursor
GitHub Copilot
Gemini CLISupported by the Atlas install prompt. Skill-specific requirements vary.
Pack stats
What’s included
Sindri Finalize
Auto-invoked by sindri-loop on clean termination with ≥1 kept commit. Generates the PR body, pushes the branch, opens a PR via gh, surfaces the URL, and archives the current run. A manual /sindri:finalize (advanced/debug) can force a PR from current kept commits; otherwise do not invoke directly.
Sindri Loop
The sindri orchestrator. Runs one experiment per invocation. Reads state, checks the HALT sentinel and termination predicates, picks next candidate, dispatches an experiment subagent via the Task tool, acts on the result (commit or reset), records to state, and schedules the next wakeup — unless terminated. Invoked automatically via ScheduleWakeup during an active run. A single manual tick is available via /sindri:loop (advanced/debug); otherwise do not invoke directly.
Sindri Scaffold Benchmark
Scaffold .claude/scripts/sindri/benchmark.py by scanning the repo, proposing a measurement in natural language, translating the user's refinements into Python, and running the script once to verify it emits a valid METRIC line. Never ask the user for Python or shell syntax.
Sindri Start
Interactive, one-shot scaffolding for a new sindri optimization run. Parses goal, ensures benchmark.py exists, scans repo for candidates, gets pool approval, calls sindri init (which creates the branch, runs baseline, seeds state), and schedules the orchestrator. Use when the /sindri:forge command receives a new goal statement.
Quick start
- Choose your skills. Select individual skills above, or choose the entire pack.
- Copy the install prompt. Use Copy install prompt above the list.
- Send it to your agent. Paste into your coding agent’s chat. Your agent reviews the source and guides the setup.
Try it with your agent
After installation, start with a request like:
Use the Sindri Finalize skill to help with my task.
Ask me for the context you need before getting started.Requirements & license
- A coding agent that can access your project files and install skills.
- Your agent checks each skill’s instructions, dependencies, and permissions during setup.
- License details are not provided in these listings. Check the author’s terms before using the skills.