← 返回 Skills 市场

continuity-kernel

Name: continuity-kernel
Author: tkuehnl

作者 Todd Kuehnl · GitHub ↗ · v0.2.0

cross-platform ✓ 安全检测通过

445

总下载

当前安装

版本数

在 OpenClaw 中安装

/install continuity-kernel

功能描述

OpenClaw continuity kernel for fail-open llm_input injection, deterministic runtime contracts, and shadow-mode eval receipts.

使用说明 (SKILL.md)

When to use this skill

Use this skill when the user wants to:

Inject bounded continuity context into llm_input without manual configuration
Persist Soul Card + Mission Ticket state with deterministic runtime contracts (durable default SQLite path, zero-config)
Keep agent behavior fail-open if continuity logic or storage has an error
Generate reproducible runtime proof receipts for audits and reviews
Run shadow-mode evals and append chunk-level quality snapshots

Commands

# Run full test suite
python3 -m unittest discover -s tests/continuity-kernel -p 'test_*.py'

# Generate deterministic runtime contract proof artifact
python3 skills/continuity-kernel/generate_runtime_contract_proof.py

# SC-02 core-shadow gate (trace optional, synthetic allowed for this suite)
python3 skills/continuity-kernel/run_shadow_eval.py --layer 'Soul Card' --chunk 'SC-02' --suite core-shadow --runs 100 --append artifacts/continuity-kernel/p0-evals.json

# SC-02 perturb robustness gate (trace-backed, deterministic selector + size_only)
python3 skills/continuity-kernel/run_shadow_eval.py --layer 'Soul Card' --chunk 'SC-02' --suite memoryarena-mini-perturb --perturb-profile deletion --selector-mode deterministic --compaction-policy size_only --runs 100 --trace-jsonl tests/continuity-kernel/fixtures/shadow_eval_trace_memoryarena_mini_perturb.jsonl --append artifacts/continuity-kernel/p0-evals.json

# SC-02 perturb robustness gate (trace-backed, dual-route + attention-preserving)
python3 skills/continuity-kernel/run_shadow_eval.py --layer 'Soul Card' --chunk 'SC-02' --suite memoryarena-mini-perturb --perturb-profile deletion --selector-mode dual_route_experimental --compaction-policy attention_preserving_experimental --runs 100 --trace-jsonl tests/continuity-kernel/fixtures/shadow_eval_trace_memoryarena_mini_perturb.jsonl --append artifacts/continuity-kernel/p0-evals.json

Options

run_shadow_eval.py options:

--layer TEXT — Active P0 layer name
--chunk TEXT — Active chunk id/name
--suite core-shadow|memoryarena-mini|memoryarena-mini-perturb — Task suite for eval scoring
--perturb-profile none|deletion|noise_injection|reorder — Deterministic perturb profile for memoryarena-mini-perturb
--selector-mode deterministic|dual_route_experimental — Selector mode tag for A/B eval receipts
--compaction-policy size_only|attention_preserving_experimental — Compaction policy label for dropped-field A/B eval comparison
--runs N — Number of evaluated shadow runs (default: 100)
--trace-jsonl PATH — JSONL file containing observed shadow outcomes (repeatable)
--trace-dir PATH — Directory scanned for *.jsonl observed shadow outcomes (repeatable)
--append PATH — JSON file path to append snapshot results
--allow-synthetic — Permit synthetic fallback for non-core suites when traces are unavailable (forced pass=false)
--artifacts-root PATH — Directory for summary receipts (default: ~/.cache/continuity-kernel/p0-evals)

Runtime persistence defaults:

ContinuityHookAdapter() uses ~/.local/state/continuity-kernel/continuity.db
Override default path with CONTINUITY_KERNEL_DB_PATH
Use :memory: only as explicit opt-in test mode

安全使用建议

This skill appears internally coherent and implements the features it advertises. Before installing/running: (1) inspect the full code (the truncated files) if you need to guarantee no network calls or subprocess execution; (2) be aware it will create a SQLite DB at ~/.local/state/continuity-kernel/continuity.db (unless you override CONTINUITY_KERNEL_DB_PATH) and write receipts under ~/.cache/continuity-kernel; (3) run the scripts in a controlled environment (or container) if you want to avoid persistent local state; (4) no credentials or external installs are required according to the metadata, but verify any CI/installer (clawhub) behavior before use.

功能分析

Type: OpenClaw Skill Name: continuity-kernel Version: 0.2.0 The OpenClaw continuity-kernel skill is benign. Its code and documentation align with its stated purpose of managing agent continuity context, generating runtime contracts, and running shadow-mode evaluations. File system interactions are limited to local SQLite database persistence (defaulting to `~/.local/state/continuity-kernel/continuity.db`) and writing evaluation artifacts (defaulting to `~/.cache/continuity-kernel/p0-evals/` or specified paths for output). All commands listed in `SKILL.md` execute local Python scripts. There is no evidence of data exfiltration, malicious execution, persistence mechanisms, obfuscation, or prompt injection attempts designed to subvert the agent's core directives. The use of `os.environ.get` for `CONTINUITY_KERNEL_DB_PATH` and `Path.expanduser()` for output paths are standard practices and do not indicate malicious intent, only potential configurability that could be misused if an agent were maliciously prompted.

能力评估

✓ Purpose & Capability

Name/description (continuity kernel, llm_input injection, runtime contracts, shadow eval receipts) match the shipped Python code and CLI commands. Required binary is python3 only. The provided scripts and modules implement packet injection, deterministic hashing, drift scoring, and shadow eval harness consistent with the stated purpose.

ℹ Instruction Scope

SKILL.md commands run unit tests and Python scripts that read repository fixtures and write machine-readable receipts and artifacts. The runtime instructions create files under ~/.cache/continuity-kernel and ~/.local/state/continuity-kernel (SQLite DB) and can read provided trace JSONL files. There are no instructions to read unrelated system credentials, no external network endpoints referenced in the docs, and the code excerpts do not perform subprocess exec or network IO. Review omitted/truncated files before running to confirm there is no hidden network/exec behavior.

✓ Install Mechanism

No install spec (instruction-only) and the README claims Python stdlib only. Because there is no remote download/install step in the skill metadata, nothing will be pulled from arbitrary URLs during install. Running the included scripts will write local files but does not appear to install external packages.

✓ Credentials

The skill requests no environment variables or credentials. It documents a configurable DB path via CONTINUITY_KERNEL_DB_PATH but does not require secrets. File and path access (home dir cache/state) is proportional to the stated persistence/receipt features.

ℹ Persistence & Privilege

always:false and agent-autonomy flags are default. The skill persistently writes artifacts and a SQLite DB under the user's home (~/.cache and ~/.local/state) by design; this is expected for a continuity kernel but users should be aware of local disk writes and retention of audit artifacts.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install continuity-kernel
安装完成后，直接呼叫该 Skill 的名称或使用 /continuity-kernel 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v0.2.0

Anvil AI rebrand + labs.anvil-ai.io normalization.

元数据

Slug continuity-kernel

版本 0.2.0

许可证 —

累计安装 0

当前安装数 0

历史版本数 1

常见问题