migrate-mem
Skill11 starsChanged 12 days agoMITView on GitHub
Establish useful memory for a project that already has development or research history, or improve an incomplete memory setup. Survey the project and its existing evidence, then initialize or enrich human-readable StrictDoc docs/ with current norms and historical context. Use Magic Context and other available memory alongside it where useful. No previous workflow, legacy folder convention, or pre-existing StrictDoc skeleton is assumed.
In this repository39 other files it gives its agents.
- accel-initSkill
- accel-platform-installSkill
- accel-pytorch-code-portingSkill
- accel-pytorch-pythonSkill
- accel-skill-templateSkill
- model-datasetsSkill
- platform-environment-skill-auditSkill
- agent-reach-installSkill
- code-reviewSkill
- consolidatorSkill
- diagnosing-bugsSkill
- distill-project-kbSkill
- domain-modelingSkill
- editable-vector-slidesSkill
- goal-firstSkill
- grillingSkill
- image-genSkill
- load-memSkill
- personal-ui-tasteSkill
- plan-briefSkill
- project-profileSkill
- quick-doSkill
- readable-docsSkill
- save-memSkill
- steady-doSkill
- tddSkill
- teachSkill
- writing-skillSkill
- kb-8bit-quantizationSkill
- kb-accel-sdkSkill
- nju-pptSkill
- exp-batchSkill
- exp-campaignSkill
- exp-discussSkill
- exp-probeSkill
- exp-workflowSkill
- kernel-workerSkill
- pytorch-perf-tuningSkill
- research-evidenceSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Use judgment about investigation depth, document structure, storage overlap, and which questions need the user's input.
- Distinctive: line 18
Read the useful sources in context rather than assuming particular note names, plan directories, or an older development methodology.
- Distinctive: line 41
Keep existing source material and working documentation useful.
- Very specific: line 43
Before drafting or revising prose, load readable-docs for the shared project writing style: phrases, lists, and compact tables without …
- Distinctive: line 49
Use the available StrictDoc installation or project environment.
- Distinctive: line 51
Consider whether a new reader can tell what applies now, why it applies, what happened before, and what remains uncertain.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Migrate Memory
Intent
Help an existing project acquire a readable, trustworthy memory foundation. The emphasis is understanding the project and initializing useful StrictDoc documentation, not moving old files or converting every artifact.
Use judgment about investigation depth, document structure, storage overlap, and which questions need the user's input. There is no fixed inventory, mandatory classification report, or approval ceremony for every initialization.
Understand the project as it exists
- Look across project instructions, README, relevant implementation, research notes and history. Understand purpose, architecture or research question, current state and unfinished work. For research, recover the major routes and their evidence from existing derivations, observations and results, not only code; organize them through
project-profileandsave-mem, without choosing a sole winner or inventing missing evidence. - Existing Magic Context memories and claude-mem observations can recover decisions or experiences not visible in code. Their absence does not prevent initializing StrictDoc.
- Read the useful sources in context rather than assuming particular note names, plan directories, or an older development methodology. Generated output and exhaustive transcripts rarely need the same attention as key code and decisions.
- Separate intended policy, observed implementation, past experience, and unresolved questions. Newer evidence can reveal change without automatically settling what the current rule ought to be.
Build a useful memory foundation
The companion save-mem skill describes memory maintenance and StrictDoc mechanics. Its assets/docs-skeleton/ is a starting point, located through the skill registry or the sibling ../save-mem/ directory, not a fixed home-directory path.
| Common location | Useful content |
docs/project_memory/decisions.sdoc | Current rules and decisions, their scope and rationale, and retired predecessors |
docs/project_memory/journal.sdoc | A dated project baseline, important experiences, progress, and open questions |
docs/handbook/ | Reference material worth maintaining: architecture, usage, specs, or operational guidance |
docs/strictdoc_config.py | The StrictDoc project configuration |
- Adapt this layout to what already exists. Missing docs or a missing skeleton are normal starting conditions; an existing docs directory is not an invitation to overwrite it.
- Establish a single current project profile with
project-profile, reusing this survey and any equivalent existing overview. The StrictDoc default isdocs/handbook/project-profile.sdoc; register its actual path in the project entry soload-memreads it first. Follow that skill for content and verification rather than duplicating a template here. - Distill enough knowledge to make the project understandable. Prefer a small useful baseline over a document for every module or every past action.
- Make current StrictDoc norms the source of truth, distinguishable from historical journal entries and reports. Statuses such as Active, Proposed, Deprecated, and Superseded help preserve this distinction as the project evolves.
- When sources conflict or the intended rule is unclear, surface the consequential question. Avoid inventing a rationale or presenting inferred policy as confirmed. If a settled decision has genuinely changed, maintain the current norm and briefly tell the user rather than waiting for a separate documentation command; ordinary memory and explanatory cleanup can be more flexible.
- The same fact or decision may also be stored in Magic Context. Choose overlap that improves recall and resilience; neither one-store-per-fact nor mirror-everything is the goal.
Preserve the project while improving readability
- Keep existing source material and working documentation useful. Add or integrate rather than automatically moving, archiving, deleting, or reformatting everything; discuss consequential restructuring with the user.
- On later runs, build on existing nodes and references instead of recreating the baseline. Observe the project's history-preservation and version-control conventions.
- Before drafting or revising prose, load
readable-docsfor the shared project writing style: phrases, lists, and compact tables without losing facts or conditions. It applies beyond memory anddocs/as well. - Where helpful, connect conclusions to source files, commits, or memory references. Original documents can remain as evidence rather than being copied verbatim into the new handbook.
- Startup/save guidance in
AGENTS.mdcan help future Agents find the memory. Align it with the agreed intent where needed, preserving unrelated project instructions.
Check the result in the real project
- Use the available StrictDoc installation or project environment. Inspect existing config and grammar before changing them; avoid assuming a specific version, environment manager, or plugin installation.
- Validate
.sdocand configuration changes withstrictdoc export .from the docs root, and inspect enough of the export to see whether lists, tables, and notes are readable. - Consider whether a new reader can tell what applies now, why it applies, what happened before, and what remains uncertain. A successful export alone does not establish that the content is correct.
- Briefly report the useful memory established, unresolved questions, and actual validation status. Missing integrations or unverified runtime claims can remain explicit gaps rather than reasons to invent completion.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

