Cognitive Deadlift

    1

    Keep developer judgment active with 21 evidence-based workflows for AI-assisted coding.

    Free

    1 useSecurity scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Cognitive Deadlift

    Example session with this skill installed

    Use Cognitive Deadlift release-readiness on this source checkout. Verify the skill catalog, ownership metadata, license notices and actual validation results before packaging it.

    • Read your context and instructions
    • Compiled the cognitive deadlift

    Actual source-check result

    The Cognitive Deadlift source checkout passed its repository gate. This transcript is from the actual validation command; it is not a model-behavior benchmark.

    .venv/bin/python scripts/validate_repo.py
    validate_repo: ok
    .venv/bin/python scripts/validate_skills.py
    validate_skills: ok
    .venv/bin/python scripts/validate_policies.py
    validate_policies: ok
    .venv/bin/python scripts/validate_harnesses.py
    validate_harnesses: ok
    .venv/bin/python scripts/validate_context_packs.py
    validate_context_packs: ok
    .venv/bin/python scripts/validate_skills.py --slop-only
    slop_scan: ok
    .venv/bin/python scripts/grade_skills.py --min-score 100
    
    # Skill Contract Report
    
    Static packaging checks only; scores do not measure agent performance.
    
    | Skill | Overall | Score | Lowest Axis | Finding |
    | --- | --- | ---: | --- | --- |
    | `agent-security-boundary` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `alternatives-before-code` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `assumption-audit` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `complexity-budget` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `debugging-lab-notebook` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `diff-interrogation` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `docs-claim-audit` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `evidence-to-test` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `explain-without-ai` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `failing-test-first` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `god-mode` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `problem-framing` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `read-the-docs-first` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `release-readiness` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `runtime-adapter-smoke` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `skill-authoring-gate` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `skill-deprecation-review` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `skill-overlap-audit` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `thinking-ledger-review` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `trace-the-code` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    | `transcript-review` | A+ | 100.0 | Metadata (A+) | Named skill with a concise description (at most 240 characters). |
    
    
    ## Detailed Axis Scores
    
    ### agent-security-boundary
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### alternatives-before-code
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### assumption-audit
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### complexity-budget
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### debugging-lab-notebook
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### diff-interrogation
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### docs-claim-audit
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### evidence-to-test
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### explain-without-ai
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### failing-test-first
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### god-mode
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### problem-framing
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### read-the-docs-first
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### release-readiness
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### runtime-adapter-smoke
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### skill-authoring-gate
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### skill-deprecation-review
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### skill-overlap-audit
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### thinking-ledger-review
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### trace-the-code
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    ### transcript-review
    
    | Axis | Grade | Score | Finding |
    | --- | --- | ---: | --- |
    | Metadata | A+ | 100 | Named skill with a concise description (at most 240 characters). |
    | Evidence contract | A+ | 100 | Scope, workflow, evidence and action/data boundaries contain guidance. |
    | Package integrity | A+ | 100 | Metadata, examples, local links and content checks pass. |
    | Routing review cases | A+ | 100 | Distinct positive and negative requests with reasons are present. |
    | Context cost | A+ | 100 | Root stays within the 450-word review budget; details can load on demand. |
    
    
    .venv/bin/python -m ruff check .
    All checks passed!
    .venv/bin/python scripts/security_scan.py
    security_scan: ok
    .venv/bin/python scripts/doctor.py
    Cognitive Deadlift doctor
    Overall: ready
    - ok: README (README.md)
    - ok: skills index (skills_index.json)
    - ok: skills directory (skills)
    - ok: tests directory (tests)
    - ok: Makefile (Makefile)
    - ok: CI workflow (.github/workflows/ci.yml)
    - ok: harnesses (harnesses)
    - ok: context packs (context-packs)
    - ok: thinking budget (policies/thinking-budget.yaml)
    - ok: skill atrophy taxonomy (docs/skill-atrophy-taxonomy.md)
    - ok: AI slop taxonomy (docs/ai-slop-taxonomy.md)
    - ok: harness validator (scripts/validate_harnesses.py)
    - ok: context pack validator (scripts/validate_context_packs.py)
    - ok: policy validator (scripts/validate_policies.py)
    .venv/bin/python -m pytest
    ...................................................                      [100%]
    51 passed in 3.89s
    

    Ownership and distribution checks

    The built wheel names Alinafe Matenda as author and maintainer. Its license expression is MIT AND CC-BY-NC-SA-4.0, and LICENSE, LICENSE-CONTENT and NOTICE are embedded byte-identically. The README and marketplace entry include the same author/copyright attribution. Original skill bodies and licenses remain unchanged. The marketplace ZIP is the runtime/content edition; use a full source checkout for the validation commands above.

    Remaining limits

    These checks validate repository structure, fixtures and packaging. They do not prove every agent follows the instructions, replace human review, or provide a broader commercial content license. Marketplace acceptance is separate from these local checks.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Created and maintained by Alinafe Matenda. Copyright (c) 2026 Alinafe Matenda.

    Cognitive Deadlift is a skill-retention harness for developers using AI coding assistants. It keeps problem definition, code reading, testing, tradeoff analysis and review judgment in the loop.

    What you get

    • 21 existing workflow skills, with examples and routing/review fixtures.
    • Risk-based thinking-budget policy, four context packs and five review harnesses.
    • A small entry point that selects the relevant workflow instead of loading every skill for every task.
    • Source attribution and the original license notices.

    Use problem-framing for an unclear request, trace-the-code before changing an execution path, failing-test-first to prove a behavior, and diff-interrogation before accepting a change. explain-without-ai asks the human to explain the mechanism; it does not fabricate evidence of understanding.

    Package and validation scope

    This is the runtime skill/content edition. Repository validation tooling is available in the full source checkout, not this runtime bundle. The source's full gate passes: 51 tests, skill/catalog/policy/harness/context validators, lint and security hygiene; distribution build and RepoWave checks also passed. These are repository and fixture results, not a benchmark proving every model follows instructions.

    An agent needs access to the bundled references and the user's authorized project files. Codex, Claude and Gemini adapter metadata are included. Full live compatibility across every agent is not certified. No external account, API key or paid service is required by the core workflow. Human code review remains necessary.

    Ownership and license

    Cognitive Deadlift is free to use under its existing split license. Code and adapter manifests: MIT. Skills, documentation and written methodology: CC BY-NC-SA 4.0. Preserve attribution, indicate adaptations and retain share-alike terms. Commercial use of the content requires separate permission from the maintainer. No broader commercial content license is included here.

    Original source and license records: https://github.com/alinafe82/cognitive-deadlift Maintainer: https://github.com/alinafe82

    Runtime/content edition, not the repository validation toolchain. Cannot guarantee that a model follows instructions or that a developer understands a change. Human review is still required. No full live behavior benchmark across all agents is claimed. Content remains CC BY-NC-SA 4.0; commercial use requires separate permission.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    1 use

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 27 days ago

    • Free with an Agensi account

    Requires

    Python

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Browser
    Network Access

    Reads relevant source, diffs, tests and selected bundled references. Shell checks and project edits are used only when the requested task requires them and access is authorized. Browser/network access is optional for documentation or task-specific evidence. No credential collection, background account access, automatic dependency installation or telemetry.

    Listed27 days ago

    What's inside

    Frequently Asked Questions

    Cognitive Deadlift provides 21 structured workflows that ask for concrete evidence, such as a narrow failing test or an execution trace, before changing code. These are instructions for the assistant, not a guarantee that it will comply; a human must review its work.

    The workflows are written for general software engineering. Use them with an agent that can read the bundled references and your authorized project files. Included adapter files are structurally validated; universal compatibility across languages, frameworks, and live agent platforms has not been demonstrated.

    The free runtime ZIP includes 21 workflow skills, a routing entry, supporting references, examples, fixtures, policy, adapter files, and the original license and attribution notices. Source validation tools and tests require the full GitHub source checkout. Code and adapters are MIT; skill and other content is CC BY-NC-SA 4.0. Commercial use of that content requires separate permission.

    Import the runtime bundle into a compatible agent and make its bundled references available. The root SKILL.md routes to an appropriate workflow. Grant only the repository and tools needed for your task. Full source validation is a separate process documented in the GitHub repository.

    The tracing workflows can help investigate unfamiliar code by recording entry points, state changes, and side effects. Results depend on available source, tests, authorized tools, and human review; no effectiveness or accuracy guarantee is made.

    This is an instructional workflow bundle, not an autonomous verification service. It cannot guarantee correct code, security, or assistant compliance. Repository access and optional shell or network tools require your authorization. The published 51 checks validate the source package and its structure, not agent performance.

    New in Prompt & Skill Engineering