agentsclimarketplace
Sign in

Coordination conformance

Skill wildcard/codex-to-claude-coordinator/skills/coordination-conformance

Test and compare agent-harness coordination capabilities with privacy-safe evidence. Use to record, redact, validate, and summarize whether Codex, Claude, Cursor, Copilot, Goose, OpenHands, or another coordinator or worker can start, steer, inspect, stop, review, preserve scope, expose usage, or produce a trustworthy handoff.From its SKILL.md

Install
npx -y skills add wildcard/codex-to-claude-coordinator --skill coordination-conformance

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.8 KB, 517 tokens by cl100k_base, as published. Nobody here has run it

Coordination Conformance

Test harness capabilities through local, privacy-safe evidence. Read references/evidence-policy.md before collecting authenticated product observations.

Prepare the Run

  1. Define one falsifiable capability question and exact harmless input.
  2. Use a disposable fixture with no credentials, personal data, remotes, connectors, or unrelated transcripts.
  3. Create a run directory outside the installed skill. Keep raw captures outside the repository.
  4. Record unsupported, unobserved, and untested behavior distinctly.

Redact and Record

Run scripts/redact.py before admitting screenshots, transcripts, or logs. Image evidence requires an opaque rectangle and a .redacted derivative. Image redaction requires Python 3.9 or newer and the dependency in requirements.txt; text redaction, recording, validation, and summarization use only the Python standard library.

Build an observation body as JSON, then use scripts/record.py to append it. The recorder hashes evidence, rejects raw screenshots, scans text for common identifiers and secret patterns, and keeps timestamps append-only. Pass --privacy-checked only after manually inspecting every derivative; the recorder does not infer that review. Use --evidence-source synthetic for harmless generated fixture artifacts; redacted evidence is always recorded as a redacted-derivative.

Validate and Summarize

Run:

scripts/validate.py <observations.jsonl> --root <run-directory>
scripts/summarize.py <observations.jsonl> --run-id <run-id> --root <run-directory>

Trust a capability only when its operation has a passing observation and traceable evidence. The summarizer preserves missing operations as unknown; it never converts absence into unavailable or support.

For quota-like signals, set threshold_eligible only when the value has explicit model scope, consumed direction, a percentage, reset or window semantics, and known freshness. For stop behavior, archive or idle never passes without a terminal worker state.

Preserve the Boundary

  • Do not automate consequential approval prompts.
  • Do not retain raw screenshots in the repository.
  • Do not accept absolute or traversing evidence paths.
  • Do not record account identifiers or machine-specific home paths.
  • Do not collapse per-operation results into one compatibility score.
  • Keep vendor-specific inputs and expected states in the experiment plan, not in this skill's reusable policy.

What ships with it: 9 files

32.3 KB alongside SKILL.md, 4 of them executable

agents/

references/

schemas/

scripts/

Gives 0 of the 12 instructions most context ai engineering skills give in 517 tokens

Counted across 1,328 of the 2,349 authors here whose files we hold, read 2026-09-06

  • Dispatch a fresh subagent for each taskin 76 of 1328, across 59 files
  • Perform spec compliance review before code quality reviewin 44 of 1328, across 34 files
  • Dispatch a final code reviewer after all tasksin 38 of 1328, across 26 files
  • Answer subagent questions before allowing implementationin 36 of 1328, across 26 files
  • Use the least powerful model capable of the taskin 33 of 1328, across 26 files
  • Create a TodoWrite list for all tasksin 32 of 1328, across 22 files
  • Perform a task review after each implementationin 31 of 1328, across 24 files
  • Extract all tasks and context from the planin 29 of 1328, across 20 files
  • Provide full task text to subagentsin 28 of 1328, across 20 files
  • Use git worktrees for isolated workspacesin 25 of 1328, across 20 files
  • Specify the model explicitly when dispatching a subagentin 23 of 1328, across 18 files
  • Execute all tasks from the plan without stoppingin 21 of 1328, across 16 files

Said here and by no other author read

  • Read evidence policy before collecting observations
  • Define one falsifiable capability question
  • Use a disposable fixture without credentials
  • Create run directory outside the skill
  • Record unsupported behavior distinctly
  • Run redaction script before admitting evidence

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.