tokcount

Use this skill when the user wants to measure how many tokens are in a codebase and get a fast pricing estimate.

Preconditions

• •tokcount is installed and available on PATH.
• •You have a repository path to scan.

Standard workflow

1. •Resolve the repository path to scan (default current directory).
2. •Run summary mode first:
   • •tokcount <repo-path>
3. •If the user needs machine-readable output:
   • •tokcount <repo-path> --output json
4. •If the user needs full structure detail:
   • •tokcount <repo-path> --tree
5. •If the user requests tokenizer-specific counting:
   • •tokcount <repo-path> --tokenizer estimate
   • •tokcount <repo-path> --tokenizer openai
   • •tokcount <repo-path> --tokenizer anthropic
6. •If the user has extra excludes:
   • •tokcount <repo-path> --ignore .tokcountignore

Output expectations

Summary mode includes:

• •repository path
• •tokenizer used
• •files scanned and ignored
• •total token count and line estimate
• •top token-contributing directories (for pruning)
• •Intent Systems pricing estimate block
• •disclaimer that pricing is directional and should be confirmed with Intent Systems

JSON mode includes:

• •repository
• •tokenizer
• •total_tokens
• •total_files
• •ignored_files
• •directories[] with path, tokens, percentage
• •pricing_estimate

Reporting guidance

When presenting results back to a user:

• •lead with total tokens and the largest directory contributors
• •include the estimated Proof Pilot cost
• •call out if one directory dominates token share and suggest ignoring non-core directories
• •recommend --tree when architecture-level breakdown is needed