/claims

Claims Command

Manage the full lifecycle of sourced claims: submission, verification against cited sources, dashboard review, source inspection, user-guided resolution, and interactive cobrowsing recovery.

Usage

/claims <mode> [options]

Modes

submit — Ingest new claims

/claims submit "claim statement" --source "https://url" --title "Source Title"
/claims submit --batch   (reads claims from context or file)

Submit one or more claims for tracking and future verification.

Required for single claim:

• First argument: the claim statement (quoted)
• --source: source URL
• --title: human-readable source title

Batch mode:

• --batch: submit multiple claims from the current conversation context or a provided JSON array
• Each batch item needs: statement, source_url, source_title

Processing:

1. Parse the claim(s) from arguments
2. Invoke the cogni-claims:claims skill with mode submit
3. Report how many claims were submitted and their IDs

verify — Run verification pipeline

/claims verify
/claims verify --id claim-abc123

Verify unverified claims against their cited sources. Optionally re-verify a specific claim.

Options:

• --id <claim-id>: verify (or re-verify) a specific claim
• No options: verify all unverified claims

Processing:

1. Invoke the cogni-claims:claims skill with mode verify
2. The skill dispatches claim-verifier agents in parallel (one per unique source URL)
3. Present verification summary to user

dashboard — Show claims overview

/claims dashboard
/claims dashboard --show-resolved

Display all claims grouped by status with actionable information.

Options:

• --show-resolved: include resolved claims in the display

Processing:

1. Invoke the cogni-claims:claims skill with mode dashboard
2. Render the formatted dashboard

inspect — Review claim evidence

/claims inspect <claim-id>

Show detailed deviation evidence for a specific claim, with option to open the source in the browser.

Required:

• <claim-id>: the claim to inspect

Processing:

1. Invoke the cogni-claims:claims skill with mode inspect
2. Display deviation details with source excerpts
3. Offer browser inspection if deviations exist

resolve — Resolve a deviated claim

/claims resolve <claim-id>

Present resolution options for a claim with detected deviations.

Required:

• <claim-id>: the claim to resolve (must have status deviated)

Processing:

1. Invoke the cogni-claims:claims skill with mode resolve
2. Present the claim, deviations, and evidence
3. Show resolution options via AskUserQuestion
4. Record the user's decision

cobrowse — Interactive source recovery

/claims cobrowse
/claims cobrowse --id claim-abc123
/claims cobrowse --url "https://specific-source.com"

Open source URLs in the user's browser for interactive recovery of source_unavailable claims. The user helps navigate authentication, cookie banners, and dynamic content while Claude reads and verifies.

Options:

• No options: recover all source_unavailable claims
• --id <claim-id>: recover a specific claim
• --url <url>: recover all claims citing a specific source URL

Requires: claude-in-chrome (user's Chrome browser with the extension active)

Processing:

1. Invoke the cogni-claims:claims skill with mode cobrowse
2. The skill opens each source URL in a new Chrome tab
3. User assists with navigation (login, cookie dismissal, scrolling)
4. Claude extracts content and verifies claims inline
5. User confirms results per source before they're saved

Examples

Example 1: Submit a single claim

/claims submit "The global AI market will reach $1.8T by 2030" --source "https://example.com/report" --title "AI Market Forecast"

Example 2: Verify all pending claims

/claims verify

Example 3: View the dashboard

/claims dashboard

Example 4: Inspect a specific claim

/claims inspect claim-abc123

Example 5: Resolve a deviation

/claims resolve claim-abc123

Example 6: Recover all unavailable sources interactively

/claims cobrowse

Example 7: Recover a specific source URL

/claims cobrowse --url "https://jasper.ai/pricing"

Working Directory

The command determines the working directory in this order:

1. --dir <path> if provided
2. Current working directory

Claims state is stored in {working_dir}/cogni-claims/.

Integration

Skill used: cogni-claims:claims (main orchestrator) Agents used (via skill): cogni-claims:claim-verifier, cogni-claims:source-inspector

Execution pattern: Command → parses mode and args → invokes claims skill → skill dispatches agents → results returned to user

Error Handling

• No claims.json found: Initialize workspace automatically
• Invalid claim ID: Report "Claim not found" with suggestion to run /claims dashboard
• No unverified claims: Report "All claims are already verified"
• Invalid mode: Show usage help with available modes