Proposal: Improving the Tekton CLI (tkn) Using Lessons from Cloudflare’s cf CLI
Document Status: Draft Design Proposal
Target Audience: Tekton CLI maintainers, contributors, and anyone interested in modern, agent-friendly CLIs
Based on: Analysis of Cloudflare’s cf CLI design (2026) and current state of tektoncd/cli
Plan: Stats gathering deep review, review round 2 (Copilot-approved)
Scope: Session recording → upload → aggregation → rollup → snapshot merge across liseur-sync and liseur Android
Explicitly out of scope: Period-over-period comparison work; Phase C (ADR) items
Branch state (uncommitted at time of writing):
main, 12 files changed + 1 new file (internal/insights/diagnostics.go)main, 6 files changed
| diff --git a/pkg/adapter/adapter.go b/pkg/adapter/adapter.go | |
| index b38efff19..89cd3a8d0 100644 | |
| --- a/pkg/adapter/adapter.go | |
| +++ b/pkg/adapter/adapter.go | |
| @@ -173,13 +173,14 @@ func (l *listener) Start(ctx context.Context) error { | |
| func (l listener) handleEvent(ctx context.Context) http.HandlerFunc { | |
| return func(response http.ResponseWriter, request *http.Request) { | |
| - var logger *zap.SugaredLogger | |
| - start := time.Now().UnixMilli() |
| #!/usr/bin/env bash | |
| # | |
| # Reproduce the resolver goroutine leak fixed by tektoncd/pipeline#10098. | |
| # | |
| # The default mode creates many PipelineRuns whose TaskRef is resolved by the | |
| # git resolver against an in-cluster HTTP server that accepts requests and never | |
| # replies. With a short git resolver fetch-timeout, the resolver reconciler's | |
| # timeout case wins, then the worker goroutine blocks forever while sending on | |
| # an unbuffered result channel in vulnerable builds. | |
| # |
| #!/usr/bin/env -S uv run --script | |
| # /// script | |
| # dependencies = ["psycopg2-binary"] | |
| # /// | |
| """Migrate atuin server.db (SQLite) -> PostgreSQL.""" | |
| import sqlite3 | |
| import sys | |
| import uuid | |
| import psycopg2 |
Title
Implement GraphQL batch fetching for .tekton files
Type Feature
Description
Introduce support for fetching multiple .tekton directory files using the GitHub GraphQL API instead of individual REST API calls. This change enables batching of file retrieval into a single request, reducing API usage and improving performance.
hack/generate-releaseyaml.sh lines 4–6: Updated TARGET_REPO_CONTROLLER,
TARGET_REPO_WATCHER, TARGET_REPO_WEBHOOK defaults to ghcr.io/tektoncd/...hack/second-controller.py lines 52 & 91: Updated help text and
--controller-image default to ghcr.io/tektoncd/...
When analyzing large codebases or multiple files that might exceed context limits, use the Gemini CLI with its massive
context window. Use gemini -p to leverage Google Gemini's large context capacity.
Use the @ syntax to include files and directories in your Gemini prompts. The paths should be relative to WHERE you run the
gemini command:
| function z() { | |
| if [[ -n ${1} ]]; then | |
| zellij attach -c ${1} | |
| else | |
| local session | |
| nmb=$(zellij list-sessions -n|grep -cv EXITED) | |
| [[ ${nmb} == 0 ]] && return | |
| (( nmb += 2 )) | |
| session=$(zellij list-sessions -r|grep -v EXITED | fzf --ghost "Zellij Session" --ansi -0 --height=${nmb} || return) | |
| [[ -n ${session} ]] && zellij attach -c "${session%% *}" |
Investigation Plan: E2E Test Failure - TestGithubPushRequestGitOpsCommentCancel
Problem Summary
The E2E test TestGithubPushRequestGitOpsCommentCancel is failing after 111.40 seconds. The test exercises the GitOps comment-based cancellation feature for push events.
Failure Timeline