project-status Skill

Reports implementation status of project components with evidence from transcripts, local repositories, GitHub, and Linear.

Overview

This skill provides accurate project status by distinguishing between what was discussed versus what is actually implemented. It combines evidence from four sources:

1. Transcripts - Meeting notes, voice memos, messages
2. Local repositories - Code files, tests, git history
3. GitHub - PRs, issues, CI/CD status, deployments
4. Linear - Issue tracking status, workflow states

Status Levels

Status	Definition	Evidence Required
Triage	Mentioned but needs review	Transcript/message reference
Backlog	Accepted, prioritized for future	Decision documented, or Linear/GitHub Issue
Todo	Scheduled, ready to start	Assigned in Linear or explicitly scheduled
In Progress	Implementation started	Code in repo or open PR
In Review	PR open, awaiting review	Open PR with review requested
Done	Working implementation	Code + tests, or merged PR with CI passing
Shipped	Deployed to production	Deployment evidence from GitHub
Canceled	Explicitly closed	Issue/PR closed as won't fix, duplicate, etc.
Unknown	Insufficient evidence	N/A

Testing the Skill

Prerequisites

1. A context database directory with:

   • data/ folder containing transcripts
   • .entourage/repos.json (optional, for code verification)

2. Repository configuration (if testing code verification):

   {
     "github": {
       "defaultOrg": "my-org"
     },
     "linear": {
       "teamId": "TEAM",
       "workspace": "my-workspace"
     },
     "repos": [
       {
         "name": "my-project",
         "path": "~/code/my-project",
         "github": "org/my-project"
       }
     ]
   }
   

Running Evaluations

1. Navigate to your context database:

   cd ~/my-context   # or ~/other-context
   

2. Start Claude Code with the plugin:

   claude --plugin-dir ~/your-plugin
   

3. Test individual cases:

   • No config: Remove .entourage/repos.json, ask "What's the status of authentication?"
   • Transcript only: With transcripts but no repos.json, ask about a discussed feature
   • Full verification: With complete config, ask "Give me the project status"

Example Test Queries

Test Case	Query
Single component	"What's the status of authentication?"
Multiple components	"Status of auth, dashboard, and payments"
Full project	"Give me the complete project status"

Expected Output Format

The skill outputs a markdown table with:

• Component name
• Status (Triage/Backlog/Todo/In Progress/In Review/Done/Shipped/Canceled/Unknown)
• Evidence description
• Source (Transcripts/Local/GitHub/Linear)
• Confidence level (Low/Medium/High/Very High)

Example:

## Status: My Project

| Component | Status | Evidence | Source | Confidence |
|-----------|--------|----------|--------|------------|
| Auth | Shipped | PR #42 merged, deployed | GitHub | Very High |
| Dashboard | In Progress | TEAM-123 In Progress, code exists | Linear + Local | High |
| Payments | Todo | TEAM-456 scheduled | Linear | High |
| Analytics | Triage | Mentioned Dec 21 meeting | Transcripts | Medium |

Evaluation Logs

Test results are stored in the context database (not the plugin):

~/my-context/evaluations/project-status/
~/other-context/evaluations/project-status/

Each evaluation run produces a timestamped JSON log with pass/fail results for each test case.

Test Cases

See evaluations/evaluation.json for the full test suite covering:

• Configuration scenarios (no config, local only, GitHub only, both)
• All status levels (Discussed through Shipped)
• Edge cases (unknown components, conflicts between sources)
• Multi-component queries