Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Run full test suite · majiayu000 | DecimalAI

agent toolingby majiayu000

Run full test suite

majiayu000/continuity-kernel

Source majiayu000/claude-skill-registry ↗License MIT

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

SkillScore100Maintained

Automatic

AI Engineer pack

Last verified 8/19/2026

What it changes

Passed 21 of 21 test cases with the skill, and 5 of 21 without it.

Without skill23%

With skill100%

Checks it passed that the model alone missed

The response specifies 100
The response names Soul Card
The response includes --layer

and 20 more in the full report

17 fixed5 passed either way0 still failing0 broke

Pass-rate lift: +77 pp
Tokens: -17%
Agent turns: 0%
Response time: -74%

Model gemini-3.6-flashMeasured 8/19/2026v122 of 22 cases passed with skill · 21 comparable

Tokens, turns and time describe resource use, not a dollar price.

Use this skill

Save it into your project's skills folder. Claude Code picks it up in its next session.

curl -fsSL --create-dirs -o .claude/skills/majiayu000-continuity-kernel/SKILL.md https://app.decimal.ai/s/majiayu000-continuity-kernel/SKILL.md

Want it pinned and updatable? pip install "decimalai>=0.15", then decimalai skills add majiayu000-continuity-kernel, and start Claude Code with claude --plugin-dir "$(git rev-parse --show-toplevel)/.agents"

1 · One-time — install the CLI (docs ↗)

pip install "decimalai>=0.15"

2 · Get the skill — writes .agents/skills/majiayu000-continuity-kernel/ and pins it in .decimal/skills.lock

decimalai skills add majiayu000-continuity-kernel

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/majiayu000-continuity-kernel/SKILL.md https://app.decimal.ai/s/majiayu000-continuity-kernel/SKILL.md

Raw file: /s/majiayu000-continuity-kernel/SKILL.md ↗ · machine-readable summary at /s/majiayu000-continuity-kernel.json ↗

Safety checks: 3 of 3 passed

Code scan CleanIntent PassedContent Clean

See it in action

One of 16 test cases the skill fixed. Same model, same prompt, with and without the skill.

1 of 16

Prompt

If I run the shadow eval CLI tool without specifying the --runs option, how many evaluation runs will it execute by default?

What was checked, without and with the skill
Checked	Without skill	With skill
The response specifies 100	Missed	Passed

case-09 · gemini-3.6-flash · 8/19/2026 · answers are capped and redacted

What's inside

The instructions your model reads when the skill loads.

SKILL.md48 linessingle file

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)

Track record

Benchmark

21 / 21

test cases passed, vs 5 / 21 without

Adoption

Users: 0
Times used: 0

Safety and source details

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/3/2026

View 15 scanner checks

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

majiayu000 / claude-skill-registry

skills/agent/continuity-kernel/SKILL.md

View on GitHub ↗

License MITImported 8/3/2026Version v1

More skills from the same source and nearby in the registry, with their headline benchmark lift.