Investigating a Superpowers Session

What it does

This is a session-investigation workflow rather than an application debugger. It first scopes the complaint, verifies the identity of the requested session and locates parent and child transcripts. It then examines dimensions such as skill timing, plan adherence, repeated work, stumbles, output quality and cost. Every finding needs a path-and-line reference. Its unusually important boundary is that it reports what happened; maintainers or triagers decide whether Superpowers itself has a defect or should change.

When to use it

• Explain why an agent repeated completed tasks after a long session.
• Investigate a workflow that ignored an approved plan or appeared unusually expensive.
• Prepare evidence for a maintainer without automatically exposing an entire private transcript.

Step-by-step workflow

1. Clarify the expected behavior, observed problem, relevant session and time or turn range before broad analysis.
2. Verify exact transcript paths and identities; measure records before bounded read-only extraction.
3. Populate the case record and examine the seven source-defined analysis dimensions, retaining only cited findings.
4. Produce the structured report with evidence, observed involvement and uncertainties; leave defect diagnosis to triage.
5. Only when requested, prepare a redacted export or issue draft, audit the scrubbed material and obtain approval before archiving or posting.

Example request

Use diagnosing-superpowers to investigate the session I identified by its full path. Between turns 40 and 65, it rebuilt two already completed export tasks instead of resuming Task 3. Confirm that session from the first prompt and timestamp, preserve its files and trace the relevant ledger reads and repeated commands. Cite each finding with path:line. Produce the report locally; do not export a bundle or create a GitHub issue.

Worked example

A transcript may contain a tool result with an embedded history exceeding a megabyte on one line. Dumping that line can overwhelm the investigating session and expose unrelated content. The companion context-safety reference requires measuring files and extracting narrow fields. If repeated work begins after a missing ledger read, the report can cite that observable sequence and explain the coverage limit. It must not turn correlation into an unsupported claim that a particular skill is defective.

What you get

A local case record and structured evidence report; an optional scrubbed bundle or issue draft only under the source’s separate consent conditions.

How to evaluate the result

• The requested session and turn range are verified, not chosen solely by recency.
• Every finding and numerical claim has a traceable source or recorded calculation.
• Nothing was bundled, archived or posted beyond the owner’s explicit choices.

What you need

User-controlled access to the identified session logs, verified record semantics and the complete companion prompts, references and templates. The full workflow assumes analyst subagents and suitable local storage.

Setup and activation

The publisher documents separate installation routes for Claude Code, Codex App and CLI, Cursor and other harnesses. Use the README for your exact client; installing the whole Superpowers framework can activate additional workflows automatically. Preserve this skill’s companion files and relative references. Review and security-scan the exact selected version before installation. This directory only reviews sources; it has not installed, executed or tested this skill on those clients. This skill depends on its prompts/, references/ and templates/ directories. Review context-safety.md and redaction-policy.md before any transcript analysis.

Limits and considerations

Session transcripts can contain secrets and personal material. Scrubbing is fallible, and the source requires the owner to review files before sharing. Never edit or delete original logs. Usage counters need understood semantics before cost totals are calculated; missing fields are not zero. The workflow stops at evidence and involvement, not prescribing a Superpowers fix.

Related workflows

• Use systematic-debugging for an application defect rather than a session complaint.
• Use verification-before-completion to check that report conclusions match the cited evidence.

License and reuse

MIT, copyright 2025 Jesse Vincent. Preserve the copyright and permission notice when redistributing substantial source portions; inspect separately supplied dependencies and assets under their own terms.

Further reading

• Reviewed skill version (immutable commit)
• Publisher setup, compatibility and telemetry notes
• Reviewed MIT license