remember
Skill4 starsChanged 16 days agoMITView on GitHub
Save feature context for future sessions. Use when finishing significant work, after important discussions, or when the user says "save context". Captures decisions, corrections, architecture, and deliberately deferred work.
69 other projects use a version of this.
Adopted by
Most recently taken up by
- hydra-db/hydradb-claude-codetheir own version, first seen Nov 8, 2026
- priyansh1210/shirotheir own version, first seen Nov 8, 2026
- jsmastery-pro/jsm-hqtheir own version, first seen Nov 8, 2026
- Yeachan-Heo/oh-my-claudecodetheir own version, first seen Nov 8, 2026
- ArcticFox2029/chamnantheir own version, first seen Nov 5, 2026
- Buddhilive/buddhi-clitheir own version, first seen Nov 1, 2026
Used with
Projects with it also use agent-context-isolation, agentica-infrastructure and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
parcadeiSkillagent-context-isolation
Agent Context Isolation
Also in 3 projects that use this
parcadeiSkillagentica-infrastructure
Reference guide for Agentica multi-agent infrastructure APIs
Also in 3 projects that use this
- +10 more from parcadei
Other versions69 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 69
- github/awesome-copilot★ 41ksame name
- Yeachan-Heo/oh-my-claude-sisyphus★ 41ksame name
- Yeachan-Heo/oh-my-claudecode★ 41ksame name
- langchain-ai/deepagents★ 31ksame name
- parcadei/Continuous-Claude★ 3.9ksame name
- parcadei/Continuous-Claude-v2★ 3.9ksame name
- parcadei/Continuous-Claude-v3★ 3.9ksame name
- zereight/gitlab-mcp★ 2.1ksame name
- H-mmer/pentest-agents★ 815same name
- SparkDevNetwork/Rock★ 701same name
59 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository7 other files it gives its agents.
- artifact-conventionsSkill
- commitSkill
- discussSkill
- git-summarySkill
- merge-fromSkill
- recallSkill
- translate-filesSkill
The file
We split this skill into the rules it gives an agent
7 rules. 0 are common, 4 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 61 edited versions we read keep these.
4
Distinctive
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 17
After significant feature work (code changes, architecture, multi-step tasks)
- Distinctive: line 18
After important discussions (planning, corrections, decisions about approach)
- Distinctive: line 19
When the user explicitly asks to save context
- Distinctive: line 20
Before ending a long session where valuable context would be lost
- Very specific: line 28
IMPORTANT: If features/ does not exist, follow Saving a feature for the first time before continuing.
- Very specific: line 80
Create features/ under the memory root resolved in Context Storage.
- Very specific: line 81
Add a _template.md file with the contents below:
“Common” is measured: kept by at least 6 in 10 of the 61 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Save Feature Context
Record the decisions behind a feature so future sessions don't rederive or reverse them. This captures what code alone cannot: why an approach was chosen, what was tried and rejected, and what was deliberately left undone.
The file is a dated decision record, not a status report. Never write progress state — how far along the work is, what's in flight — it is stale by the next session. Check git log for that instead.
When to Save
- After significant feature work (code changes, architecture, multi-step tasks)
- After important discussions (planning, corrections, decisions about approach)
- When the user explicitly asks to save context
- Before ending a long session where valuable context would be lost
Context Storage
The memory root is this session's auto-memory directory — the path named in your system prompt's memory section, i.e. ~/.claude/projects/<slug>/memory/, where <slug> is the absolute working directory with every non-alphanumeric character replaced by -.
Feature context files live in <memory root>/features/, one file per feature named {feature-name}.md (kebab-case).
IMPORTANT: If features/ does not exist, follow Saving a feature for the first time before continuing.
How to Save
- Determine the feature name — use the git branch name, the primary topic, or ask the user if unclear
- Check if a context file already exists for this feature using Glob on
<memory root>/features/ - If creating a NEW file — read the template at
<memory root>/features/_template.mdand follow its structure exactly - If UPDATING an existing file — read the existing file, preserve its structure, and merge new information into the relevant sections. Don't overwrite prior context — append to it.
- Review the transcript for:
- Architecture decisions and why they were made
- Corrections and course changes (what was tried, what was wrong, what we switched to)
- User preferences or feedback that changed the approach
- Key files created or modified and their roles
- Gotchas, things that tripped us up, or things easy to get wrong
- Work deliberately deferred, and why
- Write the context file
- Update the repo's reference docs, if it has any — only when the session produced reusable system knowledge, and only in a doc that already exists. Never create one.
- Confirm what was saved — the context file, and any reference doc updated.
What to Save
- Decisions and the reasoning behind them
- Corrections: "initially tried X, switched to Y because Z"
- User feedback that changed the approach
- Key files and their roles (paths only, not contents)
- Non-obvious patterns or conventions used
- Gotchas and things that required multiple attempts
- Work deliberately deferred, and open questions
Two Destinations
A session can produce two kinds of knowledge. Save to the right place:
| Knowledge type | Where | Format |
| Session context — decisions, corrections, what was tried, what was deferred | <memory root>/features/{feature-name}.md | Follow _template.md |
| System knowledge — how something works, reusable patterns, gotchas the team should know | The repo's existing reference docs | Lean: decisions, mental models, gotchas, file pointers. No code dumps. |
A gotcha specific to this feature stays in the feature file. One about a shared system belongs in that system's reference doc, if the repo has one.
What NOT to Save
- Obvious framework conventions — anything a competent reader gets from the framework's own docs
- Full file contents — just reference paths
- Temporary debugging state
- Information already captured in existing docs
- Code snippets or config dumps in reference docs — agents can read source files directly
- Copy, translation, and config value tweaks — visible in the diff. Save only the decision behind one, if there was one.
- In-flight status: unpushed, no PR, awaiting review. It goes stale; readers can check
git log/gh pr view. A merged PR number and sha is fine as history.
Saving a feature for the first time
- Create
features/under the memory root resolved in Context Storage. - Add a
_template.mdfile with the contents below:
# Feature: {Feature Name}
## Date
DD-MM-YYYY
## Branch
{branch-name}
## Key Files
- `path/to/file` — role/purpose
## Architecture Decisions
- Decision and why it was chosen over alternatives
## Corrections & Course Changes
- What was initially tried or assumed, what was wrong, and what we switched to
- User preferences or feedback that changed the approach
## Deferred & Open Questions
- Work consciously not done, and why
- Questions left unresolved
## Patterns Used
- Non-obvious patterns, conventions, or libraries specific to this feature
## Gotchas
- Things that were tricky, easy to get wrong, or required multiple attemptsDoes 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.

