Agent skill

Coding Trace Sanitize

by uw-syfi in uw-syfi/TraceLab

Sanitize normalized coding-trace JSONL rows for public sharing.

Apache-2.0Auto-check: notes

Install Coding Trace Sanitize

skills CLI
$ npx skills add uw-syfi/TraceLab --skill coding-trace-sanitize -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install uw-syfi/TraceLab coding-trace-sanitize --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/uw-syfi/TraceLab.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/coding-trace-sanitize .claude/skills/coding-trace-sanitize && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
coding-trace-sanitize
GitHub stars
138
Token cost
~1.3k tokens
SKILL.md length
422 words
Files
2
Skills in repo
7
Repo updated
First seen
Licence
Apache-2.0

At a glance

Sanitize normalized coding-trace JSONL rows for public sharing.

  • Works in 4 steps: Work from the coding-trace repo root,… → Confirm the input is a normalized round… → If starting from local histories, run… → …
  • Preparing trace/llmroundtrace.jsonl outputs for release
  • SKILL.md covers Overview, First Steps, Main Commands and What The Sanitizer Does, plus 2 more sections
  • Calls uv and rg

What it does

Coding Trace Sanitize is an agent skill from uw-syfi/TraceLab. Sanitize normalized coding-trace JSONL rows for public sharing. Use when preparing trace/llmroundtrace.jsonl outputs for release, removing local paths, pseudonymizing user identifiers, dropping tools[].input, replacing session/round/tool/project identifiers with stable pseudorandom values, choosing deterministic versus random sanitizer seeds, or checking what sensitive fields remain after sanitizeroundtrace.py.

Its SKILL.md is about 1.3k tokens, which your agent loads only when the skill is triggered. The skill folder holds 2 other files (for example `agents/openai.yaml`).

The repository describes itself as: An open toolkit and public dataset hub for collecting, sanitizing, analyzing, and visualizing coding agent traces. The licence is Apache-2.0.

When your agent uses it

  • Preparing trace/llmroundtrace.jsonl outputs for release
  • Removing local paths
  • Pseudonymizing user identifiers
  • Dropping tools[].input

Example prompts

  • “/coding-trace-sanitize”

Requirements

  • Python 3

Workflow steps

4 steps, taken from the first numbered list in SKILL.md.

  1. Work from the coding-trace repo root, identified by pyproject.toml, README.md, and scripts/sanitize_round_trace.py.
  2. Confirm the input is a normalized round trace such as trace/llm_round_trace.jsonl, not an expanded raw example.
  3. If starting from local histories, run $coding-trace-collect first to produce normalized rows.
  4. The default output is stdout; pass -o/--output for files. Use deterministic seeds when comparing versions; use --random-seed when…

What it can do on your machine

Read from SKILL.md and the folder at commit 11b8b14. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • uv
    • rg

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md. Its commands use uv, which can reach the network depending on how they are called.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Coding Trace Sanitize loads about 1.3k tokens when it runs. Until then it costs about 110 tokens; SKILL.md has 422 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~110
When it runs · the whole SKILL.md, loaded when a task matches
~1.3k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check: notes

The automated check noted patterns worth knowing about, such as sudo or a known installer.

  • NoteRuns commands with sudoSKILL.md:48
    The all-user sudo wrapper can sanitize immediately after collection:

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from uw-syfi/TraceLab at commit 11b8b14, republished under its Apache-2.0 licence (© uw-syfi). 422 words, ~1,273 tokens.

Download SKILL.mdSave it as .claude/skills/coding-trace-sanitize/SKILL.md (or your agent's skills folder). This skill also uses 1 other file; get the full folder from GitHub.
name
coding-trace-sanitize
description
Sanitize normalized coding-trace JSONL rows for public sharing. Use when preparing trace/llm_round_trace*.jsonl outputs for release, removing local paths, pseudonymizing user identifiers, dropping tools[].input, replacing session/round/tool/project identifiers with stable pseudorandom values, choosing deterministic versus random sanitizer seeds, or checking what sensitive fields remain after sanitize_round_trace.py.

Coding Trace Sanitize

Overview

Use this skill to prepare normalized round traces for sharing. The sanitizer targets the normalized JSONL schema, not arbitrary raw Claude/Codex logs.

First Steps

  1. Work from the coding-trace repo root, identified by pyproject.toml, README.md, and scripts/sanitize_round_trace.py.
  2. Confirm the input is a normalized round trace such as trace/llm_round_trace.jsonl, not an expanded raw example.
  3. If starting from local histories, run $coding-trace-collect first to produce normalized rows.
  4. The default output is stdout; pass -o/--output for files. Use deterministic seeds when comparing versions; use --random-seed when producing a one-off public artifact and stable cross-run ids are not needed.

Main Commands

Sanitize one timestamped private collection without writing loose files at the trace/ root:

bash
collection_directory="trace/collections/<collection_id>"
uv run python scripts/sanitize_round_trace.py \
  "${collection_directory}/merged.private.jsonl" \
  -o "${collection_directory}/merged.public.jsonl"

Use a named deterministic seed:

bash
uv run python scripts/sanitize_round_trace.py \
  "${collection_directory}/merged.private.jsonl" \
  -o "${collection_directory}/merged.public.jsonl" \
  --seed my-release-v1

Use a fresh random seed:

bash
uv run python scripts/sanitize_round_trace.py \
  "${collection_directory}/merged.private.jsonl" \
  -o "${collection_directory}/merged.public.jsonl" \
  --random-seed

The all-user sudo wrapper can sanitize immediately after collection:

bash
scripts/collect_all_users_sudo.sh --sanitize

After sanitizing, use the sanitized output for public analysis and validators:

bash
uv run python artifacts/utils/trace_db.py \
  trace/collections/current/merged.public.jsonl \
  trace/collections/current/merged.public.duckdb
uv run python artifacts/run_all.py \
  --db trace/collections/current/merged.public.duckdb
uv run python validators/run_all.py \
  --db trace/collections/current/merged.public.duckdb

What The Sanitizer Does

scripts/sanitize_round_trace.py performs these transformations:

  • Replaces project, session_id, turn_id, round_id, tool_call_id, and trace_key with stable pseudorandom alternatives while preserving relationships inside the output.
  • Replaces user, user_name, and username values with stable pseudonyms while preserving distinct-user counts and grouping.
  • Removes sensitive local context keys such as cwd, home, host, hostname, file_path, filepath, repo_url, repository_url, session_file, and workdir.
  • Removes keys ending in path-like forms such as _path, _filepath, or filepath.
  • Drops tools[].input entirely.
  • Preserves tools[].input_chars, result_chars, timing, model, provider, token counts, event types, and aggregate structure.
  • Rewrites timing_events[].tool_call_id consistently with the corresponding tool id.
  • Prints a short rows=... tools=... output=... seed=... status line to stderr.
Show full SKILL.md (169 more words)Show less

What It Does Not Guarantee

Do not claim the sanitizer proves a trace is safe for every release context. It does not inspect semantic leakage in model names, visible text summaries, tool names, timing patterns, token counts, or result sizes. It also does not sanitize arbitrary raw provider logs.

If the user wants to share raw examples, inspect or create public raw examples separately using $coding-trace-raw; do not run the normalized sanitizer on expanded raw text and assume it worked.

Verification

After sanitizing, run a shape summary:

bash
uv run python artifacts/utils/trace_db.py \
  trace/collections/current/merged.public.jsonl \
  trace/collections/current/merged.public.duckdb
uv run python artifacts/trace_facts/overview_summary/analyze.py \
  --db trace/collections/current/merged.public.duckdb --json

Use targeted searches for obvious leftovers:

bash
rg -n '"(cwd|home|session_file|workdir)"' "${collection_directory}/merged.public.jsonl"
bash
rg -n '"input":' "${collection_directory}/merged.public.jsonl"

Empty search output is a useful smoke test, not a complete privacy audit. user fields are expected to remain, but their values should look like user_<hex>. When reporting results, state the sanitizer's exact scope and any remaining review risk.

For an end-to-end smoke test, materialize the sanitized file once and run the artifact and validator dispatchers against that DuckDB. The artifact dispatcher derives the local timing-fit CSV from the same database before timing analyses consume it.

© uw-syfi, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 1 other file in skills/coding-trace-sanitize of uw-syfi/TraceLab.

  • SKILL.md
  • agents/openai.yaml

Open the folder on GitHubat commit 11b8b14

Compare with similar skills

Coding Trace Sanitize next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Coding Trace Sanitize compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Coding Trace Sanitize this skilluw-syfi/TraceLab138—~1.3kAutomated safety check: NotesApache-2.0
ShareClickHouse/ClickHouse50k—~558Automated safety check: NotesApache-2.0
Observe Traceruvnet/ruflo74k—~522Automated safety check: NotesMIT
SharingBuilderIO/agent-native7.1k—~3.4kAutomated safety check: PassNone
Tracezereight/gitlab-mcp2k1 repos~412Automated safety check: PassMIT
Deslop Shared Libsgarrytan/gstack136k—~3.3kAutomated safety check: NotesMIT

Similar skills

  • Share

    ClickHouse/ClickHouse

    Share a Claude Code session to pastila.nl and return a viewer link.

    50k GitHub stars~558 tokensUpdated today
    DatabasesAuto-check: notes
  • Observe Trace

    ruvnet/ruflo

    Trace agent execution by collecting spans and building a trace tree for a task

    74k GitHub stars~522 tokensUpdated today
    Auto-check: notes
  • Sharing

    BuilderIO/agent-native

    Framework-level sharing and privacy for user-authored resources (dashboards, documents, forms, decks, etc.).

    7.1k GitHub stars~3.4k tokensUpdated today
    Auto-check passed
  • Trace

    zereight/gitlab-mcp

    Evidence-driven tracing with competing hypotheses in parallel.

    2k GitHub starsUsed in 1 repo~412 tokens
    DevelopmentAuto-check passed
  • Deslop Shared Libs

    garrytan/gstack

    Find worthwhile shared-code extractions in recent work. An agent skill from garrytan/gstack.

    136k GitHub stars~3.3k tokensUpdated today
    Writing & ContentAuto-check: notes
  • Distributed Tracing

    wshobson/agents

    Implement distributed tracing with Jaeger and Tempo to track requests across microservices and identify performance bottlenecks.

    40k GitHub starsUsed in 12 repos~527 tokens
    DevOps & CloudAuto-check passed

More from uw-syfi/TraceLab

  • Prepare, publish, refresh, or validate the TraceLab public dataset on Hugging Face under UW-SyFI/TraceLab.

    138 GitHub stars~1.4k tokensUpdated 1 mo ago
    Auto-check passed
  • Coding Trace Analyze

    uw-syfi/TraceLab

    Run, summarize, plot, validate, and export normalized coding-trace JSONL files.

    138 GitHub stars~2.5k tokensUpdated 1 mo ago
    Auto-check passed
  • Coding Trace Collect

    uw-syfi/TraceLab

    Collect, count, and extract Claude Code and Codex CLI local histories into normalized coding-trace JSONL files.

    138 GitHub stars~1.9k tokensUpdated 1 mo ago
    Auto-check: notes
  • Coding Trace Normalize

    uw-syfi/TraceLab

    Explain and work with the normalized coding-trace JSONL row format produced by extractclauderounds.py, extractcodexrounds.py, and collectllmtraces.py.

    138 GitHub stars~1.6k tokensUpdated 1 mo ago
    Auto-check passed
  • Coding Trace Raw

    uw-syfi/TraceLab

    Read and explain raw Claude Code and Codex CLI session logs in this coding-trace repo.

    138 GitHub stars~1.3k tokensUpdated 1 mo ago
    Auto-check passed
  • Tracelab Public Release

    uw-syfi/TraceLab

    Publish TraceLab snapshots from the internal repo to the public uw-syfi/TraceLab GitHub repo as clean, mergeable, incremental pull requests from a persistent public mirror branch.

    138 GitHub stars~2.7k tokensUpdated 1 mo ago
    Auto-check passed

Questions about Coding Trace Sanitize

What does Coding Trace Sanitize do?

Sanitize normalized coding-trace JSONL rows for public sharing. Coding Trace Sanitize is an agent skill from uw-syfi/TraceLab. Sanitize normalized coding-trace JSONL rows for public sharing.

When should I use Coding Trace Sanitize?

Coding Trace Sanitize fits situations like: preparing trace/llmroundtrace.jsonl outputs for release; removing local paths; pseudonymizing user identifiers; dropping tools[].input.

How do I install Coding Trace Sanitize in Claude Code?

Run `npx skills add uw-syfi/TraceLab --skill coding-trace-sanitize -a claude-code`. Or copy the skill folder (skills/coding-trace-sanitize in uw-syfi/TraceLab) into .claude/skills/coding-trace-sanitize in your project. Claude Code loads it when a task matches its description.

How do I install Coding Trace Sanitize in Codex?

Run `npx skills add uw-syfi/TraceLab --skill coding-trace-sanitize -a codex`. Or copy the skill folder (skills/coding-trace-sanitize in uw-syfi/TraceLab) into .agents/skills/coding-trace-sanitize in your project. Codex loads it when a task matches its description.

Can I use Coding Trace Sanitize in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add uw-syfi/TraceLab --skill coding-trace-sanitize -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/coding-trace-sanitize, .gemini/skills/coding-trace-sanitize, .github/skills/coding-trace-sanitize and .opencode/skills/coding-trace-sanitize in your project.

What does Coding Trace Sanitize need to run?

Going by SKILL.md and its folder, Coding Trace Sanitize needs the command-line tools its instructions call (uv and rg). Our summary lists: Python 3.

Does Coding Trace Sanitize access the network?

SKILL.md contains no URLs. Its commands use uv, which can reach the network depending on how they are called. This is read from the text; nothing was executed.

Is Coding Trace Sanitize safe to install?

Our automated static check of SKILL.md found notes only (runs commands with sudo), nothing it rates as a warning. It is not a guarantee. Review the folder before installing.

What licence does Coding Trace Sanitize use?

Coding Trace Sanitize is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Coding Trace Sanitize use?

About 1.3k tokens (SKILL.md is roughly 5.1k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Coding Trace Sanitize?

Skills that share tags, products or a category with Coding Trace Sanitize: Share (ClickHouse/ClickHouse, 50k stars), Observe Trace (ruvnet/ruflo, 74k stars), Sharing (BuilderIO/agent-native, 7.1k stars) and Trace (zereight/gitlab-mcp, 2k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Coding Trace Sanitize?

uw-syfi (a GitHub organization) maintains it in uw-syfi/TraceLab, which has 138 GitHub stars. The repository holds 7 skills in this directory. The repository was last updated on August 22, 2026.

Source: uw-syfi/TraceLab on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.