sb-ai-lab/sirin/agent
Claude Code and agent skill by sb-ai-lab/sirin
- community
Checking that the SKILL.md file is still in the repository.
Use when working in the SIRIN repository to implement, debug, review, or document hallucination and answerability detection code, including probing, judging, uncertainty estimation, feature processors, model adapters, Hydra configs, checkpoint compatibility, Streamlit UI presets, or SIRIN tests.
Install
npx vskill@latest install sb-ai-lab/sirin/agentSKILL.md
SIRIN Framework
Use this as the entry point. Load only the reference file that matches the task.
First Checks
- Work from the repo root and pin imports with
PYTHONPATH=$PWD; the shared environment may otherwise import another checkout. - Prefer existing detector, processor, adapter, logger, splitter, and config patterns before adding anything new.
- Treat vendored dependencies, config compatibility, and cache behavior as documented in ../reference/pitfalls.md.
- When changing configs or public imports, update the matching shared reference doc.
- Keep generated or captured artifacts out of commits unless they are intentional fixtures.
Reference Router
- Need architecture, dataflow, or import paths: read ../reference/architecture.md.
- Need dataclass defaults, enum names, constants, or config field ownership: read ../reference/configs.md.
- Need side selection, layer indexing, attention behavior, feature processor capability, or compression details: read ../reference/feature-processing.md.
- Need dataset shape, runnable examples, Hydra layout, save/load, context splitting, logging, or caching patterns: read ../reference/workflows.md.
- Need known failure modes, validation traps, batch-size meanings, vendored compatibility, or environment setup: read ../reference/pitfalls.md.
- Need Streamlit UI specifics: read ../ui.md.
Common Workflows
Change Detection Behavior
- Identify the family: probing, judge, or uncertainty.
- Read the relevant imports/configs in
../reference/architecture.mdand../reference/configs.md. - Trace the flow through adapter -> feature processor -> detector -> pipeline.
- Add one focused check for the behavior you changed.
- Run the narrowest relevant test command.
Touch Feature Extraction
- Read
../reference/feature-processing.mdbefore choosingside,pooling_type,layers, orattn_implementation. - Follow the cache and attention constraints documented there.
Touch Generation or API Adapters
- Keep standard sampling explicit:
max_lengthmaps to adaptermax_tokens. - Put provider-specific request extensions where
../reference/configs.mdsays they belong. - Use an integration test only when endpoint shape matters.
Touch Checkpoint Loading
- Follow the checkpoint compatibility notes in
../reference/pitfalls.md. - Keep legacy loading behavior covered by a focused test.
Minimum Verification
Set CUDA_VISIBLE_DEVICES as needed, then run PYTHONPATH=$PWD python -m pytest -q tests/ -m integration only for tests explicitly marked integration.
Security
- Security scanPassed, 100 of 100Pattern checks on the skill's files.
- Model reviewNot runA model review is scheduled for the next rescan.
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingdocs/agent/SKILL.md
Popularity
- Repository stars
- 61
- Repository forks
- 3
- 7-day trend score
- 21
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/sb-ai-lab/sirin/agent)