Agent memory blueprint
Agent Memory Blueprint: public-safe templates, validators, and skill packaging for machine-first long-running agent memory.
npx -y skills add yystudio-cyber/agent-memory-blueprintAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Design, scaffold, validate, analyze, package, and publish machine-first memory blueprints for long-running AI agents. Use when Codex needs to create or improve a canonical multi-project memory YAML file, split execute/history/update read modes, define project skill cards and detail sections, maintain append-only event ledgers, validate strict-public safety, or prepare a public GitHub Agent Skill/template without leaking private content.
SKILL.md
5.5 KB, ~1.1k tokens by cl100k_base, as published. Nobody here has run it
Agent Memory Blueprint
Treat agent memory as a bootable knowledge system, not as notes.
When to use this skill
Use this skill to:
- design a machine-first YAML memory blueprint for agents
- scaffold a new synthetic or private memory file
- validate structure and public-safety boundaries
- analyze an existing memory shape without printing detail contents
- prepare a public GitHub Skill package from a private architecture
- explain or maintain the four-layer memory model
When not to use this skill
Do not use this skill to:
- store or publish real private data
- publish real candidate, customer, resume, email, phone, profile URL, or recruiting data
- replace a database, CRM, ATS, RPA, crawler, scraper, or SaaS product
- bypass privacy review or treat
--strict-publicas a perfect leak detector
Operating model
Build four layers:
- Boot and routing:
agent_boot,read_profiles,project_skill_cards,project_index - Execution cards: compact project cards with
current_state,next_action, paths, and do-not rules - Detail sections: SOP, constraints, validation state, commands, and deeper project context
- History ledgers: append-only
calendar_events_YYYY_MMsections for reports and retrospectives
Use multi-document YAML only when helpful. Top-level keys must be unique within and across documents.
Workflow
1. Analyze shape first
Run:
python scripts/analyze_memory_shape.py /path/to/memory.yaml --json
Use the report for structural facts only: top-level keys, read profiles, project cards, detail links, calendar sections, duplicate top-level keys, and public-risk marker counts. Do not copy live project content into public templates.
2. Choose a reference
- Use
references/architecture.mdfor the four-layer model. - Use
references/schema.mdandschemas/memory.schema.jsonfor field contracts. - Use
references/update-protocol.mdwhen editing a live private memory file. - Use
references/publication.mdbefore preparing public GitHub content.
3. Scaffold
Run:
python scripts/scaffold_memory.py --output /path/to/memory.yaml --owner OWNER_NAME
The scaffold is synthetic. Replace it with real project cards only in private files outside the public repository.
4. Validate
Run:
python scripts/validate_memory.py /path/to/memory.yaml
Before publishing:
python scripts/validate_memory.py /path/to/memory.yaml --strict-public
python scripts/release_check.py
Strict-public fails likely emails, phone-like content, recruiting/profile URLs, secret-like assignments, and non-placeholder local paths such as C:\Users\real-user\..., D:\Projects\..., /Users/real-user/..., /home/real-user/..., /mnt/data/..., /workspace/..., /var/secrets/..., /opt/internal/..., or ~/private/....
Allowed placeholder path patterns include /path/to/..., /example/..., obvious values containing OWNER_NAME, PROJECT_ID, example, demo, or placeholder, and /tmp/... for tooling convenience.
5. Package
Run:
python scripts/package_skill.py --output dist/agent-memory-blueprint.zip
The package check must reject caches, .git, .env, private memory files, private/, backups/, .pyc, and nested dist/ output.
Design rules
- Keep one canonical private memory file per actual user/workspace.
- Let
read_profileschoose execution, history, or update-memory mode. - Keep
project_skill_cardsshort enough for every execution. - Put deep context in named detail sections.
- Keep calendar history outside default execution reads.
- Use
supersedesor deprecation notes for correction instead of silent deletion. - Every meaningful update should leave a concise human-readable event.
- Validation status must be truthful:
passed,partial,blocked, ornot_run.
Before publishing
- Run every YAML intended for publication through
--strict-public. - Run
python scripts/release_check.pyormake release-check. - Inspect examples manually and confirm they are synthetic.
- Inspect
dist/agent-memory-blueprint.zip. - Do not publish live private memories, backups, real project logs, credentials, profile links, resumes, or personal data.
- Do not weaken
strict-publicto make unsafe content pass.
Resources
assets/memory-template.yaml: public starter templateexamples/: synthetic public examples for productivity, software, and research workflowsschemas/memory.schema.json: baseline JSON Schema for core structurescripts/analyze_memory_shape.py: private-safe structural analyzerscripts/package_skill.py: public-safe zip packagerscripts/release_check.py: one-command release verificationscripts/scaffold_memory.py: deterministic template generatorscripts/validate_memory.py: structural and strict-public validatortests/: regression tests for validators, examples, and private-safe analysisreferences/architecture.md: architecture modelreferences/schema.md: field contracts and naming rulesreferences/update-protocol.md: live-memory editing procedurereferences/publication.md: GitHub/public-release boundary
Gives 0 of the 12 instructions most memory context skills give in ~1.1k tokens
Counted across 674 of the 847 authors here whose files we hold, read 2026-08-06
- inform the user when setup is completein 21 of 674, across 6 files
- confirm the draft with the user before writingin 21 of 674, across 6 files
- update the agent skills block in place if it existsin 21 of 674, across 6 files
- present findings to the userin 20 of 674, across 5 files
- write the three docs files from seed templatesin 20 of 674, across 5 files
- ask the user about each decision one at a timein 19 of 674, across 4 files
- edit CLAUDE.md if it existsin 18 of 674, across 3 files
- explore current repo statein 18 of 674, across 3 files
- do not overwrite user edits to surrounding sectionsin 18 of 674, across 3 files
- back up the original file before overwritingin 16 of 674, across 8 files
- keep the memory index under 200 linesin 15 of 674
- Provide actionable steps and verificationin 13 of 674, across 2 files
Said here and by no other author read
- keep top-level keys unique within and across documents
- analyze the memory shape before editing
- use references and schemas for field contracts
- replace synthetic scaffolds only in private files
- validate memory before publishing
- run strict-public validation before publication
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.