project-context
Skill2 starsChanged 6 months agoMITView on GitHub
When the user wants to create, update, or inspect project context for agent-assisted client/product work. Use when the user mentions project context, client context, decision locations, requirements, source systems, execution tracker, review rules, delivery workflow, or wants to avoid repeating foundational project information across agent tasks. Creates or updates `.agents/project-context.md`, which other Project Skills read first; records transcript locations without ingesting full transcripts by default.
33 other projects use a version of this.
Adopted by
Most recently taken up by
- shinpr/ai-coding-project-boilerplatetheir own version, first seen Nov 13, 2026
- advait/effect-durable-agenttheir own version, first seen Nov 8, 2026
- upex-galaxy/agentic-qa-boilerplatetheir own version, first seen Nov 7, 2026
- egorfedorov/claude-context-optimizertheir own version, first seen Nov 1, 2026
- ritamgh/claude-skillstheir own version, first seen Nov 1, 2026
- OpenxIntelligence/Coding-Agent-Skillstheir own version, first seen Nov 1, 2026
Used with
Projects with it also use code-review.
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.
GoogleSkillcode-review
Adversarial systems, concurrency, memory-model, and architecture code review protocol. Use this skill whenever asked to review code, PRs, diffs, or branches in Filament.
Also in 3 projects that use this
Other versions33 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 33
- shinpr/ai-coding-project-boilerplate★ 239same name
- egorfedorov/claude-context-optimizer★ 130same name
- TenureAI/PhD-Zero★ 52same name
- upex-galaxy/agentic-qa-boilerplate★ 26same name
- sergeeey/Claude-cod-top-2026★ 9same name
- perhapsspy/project-context★ 7same name
- advait/effect-durable-agent★ 4same name
- aida-core/aida-marketplace★ 4same name
- I-Onlabs/claude-code-skills★ 4same name
- IgorBabikov/Flowcairn★ 4same name
23 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 repository6 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- decision-captureSkill
- execution-traceSkill
- requirements-from-decisionsSkill
- workspace-initSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
25 rules. 0 are common, 21 are distinctive, and 4 are very specific to this project.
0
Common
Most of the 24 edited versions we read keep these.
21
Distinctive
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 18
Use this skill to capture the stable project context an agent needs before helping with delivery work:
- Distinctive: line 45
Ask what changed or what section the user wants to update.
- Distinctive: line 48
If it does not exist:
- Distinctive: line 51
Start from lightweight sources first: README, existing docs, repository metadata, issue tracker/project names, meeting summaries, …
- Distinctive: line 53
Only read full transcripts when the user explicitly asks for transcript-derived extraction, such as finding decision candidates, …
- Very specific: line 54
If information is missing, leave explicit Unknown or TBD markers instead of inventing.
- Distinctive: line 58
Do not assume the current git repository is the whole project.
- Distinctive: line 89
Ask concise questions if needed, but prefer drafting from available source material first.
- Distinctive: line 100
Default behavior: identify where transcripts live and what role they play.
- Distinctive: line 173
After creating or updating the file, summarize:
- Distinctive: line 183
Do not copy private raw transcripts or sensitive source material unless the user explicitly asks and the project permits it.
- Distinctive: line 184
Do not read full transcripts just because a transcript folder or link exists.
- Distinctive: line 185
Do not treat meeting notes as accepted decisions unless the notes say they were accepted or the user confirms.
- Distinctive: line 186
Do not invent stakeholders, trackers, repos, requirements, or approval rules.
- Very specific: line 187
If a named approver/reviewer is explicit in the user's context or source material, record that named role/person directly instead of …
- Distinctive: line 188
If a project uses a canonical decision log, point to it instead of creating a competing one.
- Distinctive: line 189
If a project does not use a formal decision log, capture the current decision source locations instead of pretending one exists.
- Distinctive: line 190
Do not assume decision logs, transcripts, or project notes belong inside the production repository.
- Distinctive: line 206
Prefer markdown tables for source systems, stakeholders, decision locations, trackers, and review rules.
- Very specific: line 207
Use stable labels such as TBD, Unknown, Accepted, Candidate, Superseded, and Blocked instead of prose-only status descriptions.
- Distinctive: line 208
Keep source links inline in the relevant row or bullet.
- Distinctive: line 209
Keep entries concise enough for another agent to parse and update safely.
- Distinctive: line 210
Keep the context file as an index and operating guide, not a source-material digest.
- Distinctive: line 211
When local skills or agent instructions are part of the workspace, list their stable path as a source system or operating source so future …
- Very specific: line 212
Avoid ambiguous requirement or trace statuses such as bare Ready when they could be mistaken for implemented or accepted state.
“Common” is measured: kept by at least 6 in 10 of the 24 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
Project Context
You help users create and maintain .agents/project-context.md, the foundational context file for a project.
This file may live in a production repository, but it does not have to. Many projects are better represented as a workspace that contains context, transcripts, notes, and one or more git repos.
Other Project Skills should read this file before capturing decisions, deriving requirements, tracing execution, or closing out delivery work.
What this skill does
Use this skill to capture the stable project context an agent needs before helping with delivery work:
- what the project is
- who it is for
- where source truth lives
- where decisions are recorded
- how requirements should be written
- where execution happens
- how review and approval work
- what delivery closeout should include
This is not a full project management system. It is a compact, strictly formatted markdown scaffold that helps agents stop losing the plot between meetings, requirements, tickets, PRs, and documentation.
Workflow
Step 1: Check for existing context
Look for:
.agents/project-context.mdIf it exists:
- Read it.
- Summarize what is already captured.
- Ask what changed or what section the user wants to update.
- Preserve existing useful context unless it is clearly superseded.
If it does not exist:
- Offer to create it.
- Start from lightweight sources first: README, existing docs, repository metadata, issue tracker/project names, meeting summaries, user-provided notes, and known source-system locations.
- Record transcript locations as source systems, but do not ingest full transcripts by default.
- Only read full transcripts when the user explicitly asks for transcript-derived extraction, such as finding decision candidates, requirements, risks, or open questions.
- If information is missing, leave explicit
UnknownorTBDmarkers instead of inventing.
Step 1.5: Choose the right workspace boundary
Do not assume the current git repository is the whole project.
Valid shapes include:
# Context inside the production repo
repo/
.agents/project-context.md
src/
# Project workspace with repo inside
client-project/
.agents/project-context.md
transcripts/
notes/
app/ # git repo
# Project workspace with multiple repos
client-project/
.agents/project-context.md
web-app/ # git repo
api/ # git repo
docs/Prefer the smallest workspace that matches how the project is actually run. If source truth lives in transcripts, email, Linear/Jira, GitHub issues, CRM notes, or external docs, point to those systems. Do not require teams to commit formal decision logs or raw meeting transcripts into production code repos.
Step 2: Gather project context
Capture only what is useful for future agent work.
Ask concise questions if needed, but prefer drafting from available source material first.
Source budget order:
- Existing
.agents/project-context.md, if present. - User-provided project summary or notes.
- README / repo metadata / package names / docs index.
- Tracker/project names and source-system locations.
- Meeting summaries or transcript summaries.
- Full transcripts only on explicit request or when the user supplies a small excerpt for this task.
Default behavior: identify where transcripts live and what role they play; do not load or summarize the full transcript corpus while creating basic project context.
Required sections:
- Project overview
- Freshness warning
- Stakeholders and roles
- Source systems
- Decision log
- Requirement format
- Execution tracker
- Repository / PR workflow
- Review and approval rules
- Delivery closeout expectations
- Execution trace rules
- Constraints and non-goals
- Project vocabulary
- Open questions
If the context is generated from local snapshots, local repo inspection, exported docs, or stale mirrors, include a ## Freshness warning that says which facts must be re-verified before status, acceptance, blocker, or delivery claims. If the project has live systems such as Linear, GitHub, Jira, email, or CI, make clear that local snapshots are routing context, not current truth.
Step 3: Write .agents/project-context.md
Use this structure:
# Project Context
Status: active
Last updated: YYYY-MM-DD
## Project overview
## Freshness warning
## Stakeholders and roles
## Source systems
| System | Role | Link / location | Notes |
|---|---|---|---|
## Decision log
## Requirement format
## Execution tracker
## Repository / PR workflow
## Review and approval rules
## Delivery closeout expectations
## Execution trace rules
- `Accepted` requires the project-defined reviewer/client approval or acceptance evidence.
- Merged PRs, closed tickets, and local implementation evidence do not by themselves prove acceptance.
- Missing ticket/PR/check/demo/review evidence should become `Unlinked`, `Needs verification`, or a visible gap in execution trace work.
- Open source dependencies or unanswered requirement questions should mark affected requirements as blocked or partial.
- Non-goals and constraints should be tracked as scope guards, not implementation tasks.
## Constraints and non-goals
## Project vocabulary
## Open questions
## Change logStep 4: Report what changed
After creating or updating the file, summarize:
- sections created or updated
- unknowns left as
TBD - any decisions that still need confirmation
- recommended next skill, if any
Source handling rules
- Preserve links and source refs when available.
- Do not copy private raw transcripts or sensitive source material unless the user explicitly asks and the project permits it.
- Do not read full transcripts just because a transcript folder or link exists. Treat transcript locations as source-system references unless transcript extraction is the task.
- Do not treat meeting notes as accepted decisions unless the notes say they were accepted or the user confirms.
- Do not invent stakeholders, trackers, repos, requirements, or approval rules.
- If a named approver/reviewer is explicit in the user's context or source material, record that named role/person directly instead of leaving a redundant
TBDquestion. If authority is only implied, keep it role-based and mark itTBD. - If a project uses a canonical decision log, point to it instead of creating a competing one.
- If a project does not use a formal decision log, capture the current decision source locations instead of pretending one exists.
- Do not assume decision logs, transcripts, or project notes belong inside the production repository.
Output standard
A good .agents/project-context.md is:
- short enough to read quickly
- specific enough to guide future agent work
- explicit about source systems
- clear about review and approval boundaries
- honest about unknowns
- useful to downstream skills
Formatting rules:
- Preserve the section headings from the template unless the user explicitly asks for a different structure.
- Prefer markdown tables for source systems, stakeholders, decision locations, trackers, and review rules.
- Use stable labels such as
TBD,Unknown,Accepted,Candidate,Superseded, andBlockedinstead of prose-only status descriptions. - Keep source links inline in the relevant row or bullet; do not collect them in a disconnected appendix.
- Keep entries concise enough for another agent to parse and update safely.
- Keep the context file as an index and operating guide, not a source-material digest.
- When local skills or agent instructions are part of the workspace, list their stable path as a source system or operating source so future agents can load the same workflow.
- Avoid ambiguous requirement or trace statuses such as bare
Readywhen they could be mistaken for implemented or accepted state; use phrases likeRequirement readyorReady for execution tracewhen needed.
Common pitfalls
- Creating a giant knowledge base instead of a compact project context file.
- Treating every note as a decision.
- Losing source refs.
- Hiding unknowns instead of marking them
TBD. - Duplicating project management tools instead of pointing to them.
- Writing generic context that could apply to any project.
- Assuming every team has a committed
/docs/decisionsfolder or formal ADR/decision-log process. - Dumping raw transcripts into
.agents/project-context.mdinstead of extracting stable context and linking to the transcript location. - Burning context on full transcript ingestion when the task only needs source locations and operating rules.
Verification checklist
Before finishing:
.agents/project-context.mdexists or the user declined creation.- Source systems are listed.
- A freshness warning is included when any current-state facts came from local snapshots, local inspection, or exported docs.
- Decision log location is explicit or marked
TBD. - Requirement format is explicit or marked
TBD. - Execution tracker is explicit or marked
TBD. - Review rules are explicit or marked
TBD. - Unknowns are visible.
- No sensitive/private raw data was added accidentally.
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.

