agentleFS
oborchersoborchers/proqi/skills/proqi-debug/SKILL.md

proqi-debug

Skill45 starsChanged 51 days agoMITView on GitHub

Diagnose a reported Proqi failure using its content-redacted local diagnostics, stable CLI errors, and documented SQLite model. Use only when the user explicitly asks to debug Proqi or prepare a Proqi issue.

The file

Show its rules and text

We split this skill into the rules it gives an agent

18 rules. About 0 are common, 12 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈12

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Run proqi --version and proqi --json capabilities.

    line 13
  2. Very specific:

    Ask before collecting local diagnostics.

    line 26
  3. Distinctive:

    Read references/storage.md only when the failure involves persistence, migrations, leases, undo history, or submissions.

    line 45
  4. Very specific:

    Prefer proqi diagnostics collect over reading JSONL segments directly.

    line 64
  5. Distinctive:

    Remove thought text, clipboard data, credentials, personal paths, session names, pane identifiers, and identifiers that are not required to …

    line 81
  6. Distinctive:

    Create issues only for confirmed non-security defects reproduced at least twice on the identified released binary from isolated QA state.

    line 95
  7. Distinctive:

    Search open and closed issues first.

    line 97
  8. Distinctive:

    Use only existing repository labels.

    line 98
  9. Distinctive:

    Do not attach or upload diagnostics, logs, screenshots, databases, or other local evidence.

    line 101
  10. Distinctive:

    Create exactly one issue for each independently confirmed root defect, verify the created issue, and record its URL in the QA report.

    line 103
  11. Distinctive:

    If any prerequisite is uncertain, keep the finding local as unconfirmed or blocked instead of publishing it.

    line 105
  12. Very specific:

    Never open a public issue for a suspected security vulnerability.

    line 108
  13. Very specific:

    Do not mutate SQLite or bypass Proqi's CLI, owner channel, or leases.

    line 115
  14. Distinctive:

    Do not emulate a semantic mutation with TUI keys.

    line 116
  15. Distinctive:

    Do not upload logs, bundles, databases, screenshots, or thought content.

    line 118
  16. Very specific:

    Do not retry a submission whose durable state is outcome_unknown.

    line 119
  17. Distinctive:

    Do not claim a fix or successful delivery without direct evidence.

    line 120
  18. Distinctive:

    Do not open, comment on, or modify a GitHub issue without explicit approval, except for new issues created within the narrow unattended QA …

    line 121

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Proqi Debug

Diagnose one reported failure safely. Begin read-only and minimize access to scratchpad content.

Workflow

  1. Run proqi --version and proqi --json capabilities. Require the supported JSON schema, doctor in data.commands, and the exact diagnostic operation before using it. If the installed version does not advertise a command, report that limitation instead of guessing. Then run the read-only health report proqi --json doctor. Preserve every doctor check ID, status, and remediation exactly. Doctor never repairs, migrates, initializes, or recreates canonical Proqi state.
  2. Record the installed version, CLI schema version, operating system, terminal, exact invocation, approximate failure time, and smallest reproduction.
  3. Preserve the exact stable error.code, error.message, and error.details from JSON commands. Do not strengthen an intermittent symptom into a confirmed defect.
  4. Ask before collecting local diagnostics. In explicitly authorized unattended QA mode, this approval is satisfied only for diagnostics produced from the run's isolated QA state. Never collect from the user's normal Proqi state in unattended mode. Use an explicit private output path that does not already exist:

    proqi diagnostics collect --output proqi-diagnostics.json
  5. Review the bundle locally before quoting or attaching it. It is designed to omit thought text, clipboard content, session names, workspace paths, pane identifiers, and raw external responses, but it retains typed identifiers and timestamps.
  6. Correlate lifecycle events, command failures, and submission transitions by time and typed identifier. Never infer delivery failure merely from a later agent readiness state.
  7. Reproduce with the smallest safe test session where practical. Do not read unrelated sessions or copy the user's live database merely for convenience.
  8. Read references/storage.md only when the failure involves persistence, migrations, leases, undo history, or submissions. For history reports, first identify the scope: Browser history owns session rename, trash, and restore; Board history owns item and transformation operations inside one session; editor history owns exact replacements for one thought.
  9. Classify the result as expected behavior, configuration or environment, unsupported integration, unconfirmed defect, confirmed defect, or security concern. State the evidence and any missing verification.

Local files

Proqi uses the operating system's native application data directory. Diagnostic segments are inside its diagnostics subdirectory, and the database is proqi.sqlite3 in the data directory.

  • macOS: ~/Library/Application Support/proqi/
  • Linux: ${XDG_DATA_HOME:-~/.local/share}/proqi/

Prefer proqi diagnostics collect over reading JSONL segments directly. Never edit the live database, its -wal or -shm files, lock metadata, backups, or diagnostic logs. Never run repair SQL against user data.

Prepare an issue

For a confirmed non-security defect, first search existing Issues at https://github.com/oborchers/proqi/issues. Then draft the exact issue title and body in chat. Include:

  • Proqi version and CLI schema version.
  • Operating system and terminal.
  • Minimal reproduction steps.
  • Expected and observed behavior.
  • Stable error code and a short content-redacted diagnostic excerpt.
  • Whether the failure is reproducible and whether data was modified or lost.

Remove thought text, clipboard data, credentials, personal paths, session names, pane identifiers, and identifiers that are not required to correlate the failure. Do not attach the diagnostic bundle by default.

By default, obtain explicit approval of the complete issue text before running gh issue create. Opening the issue is an external mutation. Report the issue URL only after the command succeeds.

An unattended QA goal may provide advance authorization to create issues without per-issue approval. Use that exception only when the goal explicitly names this skill, restricts testing to an identified released binary, requires repeated reproduction, defines a severity model, requires duplicate search, and authorizes automatic GitHub issue creation. In that mode:

  1. Create issues only for confirmed non-security defects reproduced at least twice on the identified released binary from isolated QA state.
  2. Search open and closed issues first. Do not create a duplicate.
  3. Use only existing repository labels. Do not create or edit labels, milestones, projects, repository settings, or releases.
  4. Sanitize the complete title and body locally before creation.
  5. Do not attach or upload diagnostics, logs, screenshots, databases, or other local evidence.
  6. Create exactly one issue for each independently confirmed root defect, verify the created issue, and record its URL in the QA report.
  7. If any prerequisite is uncertain, keep the finding local as unconfirmed or blocked instead of publishing it.

Never open a public issue for a suspected security vulnerability. Follow SECURITY.md and use GitHub private vulnerability reporting instead. A private security submission still requires explicit approval unless the unattended goal separately and explicitly authorizes that exact channel.

Boundaries

  • Do not mutate SQLite or bypass Proqi's CLI, owner channel, or leases.
  • Do not emulate a semantic mutation with TUI keys. Active and inactive JSON writes must converge through the same application owner and durable receipt.
  • Do not upload logs, bundles, databases, screenshots, or thought content.
  • Do not retry a submission whose durable state is outcome_unknown.
  • Do not claim a fix or successful delivery without direct evidence.
  • Do not open, comment on, or modify a GitHub issue without explicit approval, except for new issues created within the narrow unattended QA authorization above.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.