HarnessTax: How Much Does the #Harness Matter for #CodingAgents?
@frankel.wsocial.eu
@frankel.wsocial.euHarnessTax: How Much Does the #Harness Matter for #CodingAgents?
@frankel.chHarnessTax: How Much Does the #Harness Matter for #CodingAgents?
@pulseofnations.lolThe Plugin4Shell zero-click flaw let a plugin repository owner serve malicious code under a reviewed commit hash. Claude Code and Codex patched it. Copilot and Gemini CLI did not.
#AiSecurity #ClaudeCode #Codex #CodingAgents #Copilot #GeminiCli #Plugin4shell #SupplyChain
@reksas13.bsky.socialDay 71/100 of #100DaysWithAI đ
Cursor's agent docs have no error-handling section. Just a terminal that monitors output. Your error messages are the feedback loop.
đ https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #CodingAgents
đ¤ Claude Code-assisted, reviewed by me.
@wideareaai.bsky.social
@softwarearchsummit.bsky.socialCoding Agents nehmen dir Arbeit ab, aber wer behält die Architektur im Blick?
Modul AGENTA zeigt dir, wie du agentische Entwicklung integrierst & Architektur unter Kontrolle behältst.
đ https://software-architecture-camp.de/modul-agenta-architektur-fuer-agentic-software-engineering/
@wideareaai.bsky.social
@nicolaounicos.bsky.socialAI coding agents are moving beyond simple fixes.
Android Bench 2.0 tests complex, multi-day engineering tasksâand pass rates drop from ~91% to ~28%.
The next challenge isnât writing code. Itâs completing the job.
@tmtabor.ioOne thing that Anthropic got very right is the automatic prompt to generate a CLAUDE.md file for project context. This is a smart move that enforces some of the best practices I had already been following. #CodingAgents
@o-zehentleitner.bsky.socialYour repository already is project memory.
Code, docs, tests and Git preserve the *what* and *how*.
Whatâs usually missing is the *why*:
why this design, what was tried, what was rejected.
That belongs in the repo too.
@softwarearchsummit.bsky.socialCoding Agents nutzen â und Architektur, Qualität und Kontrolle behalten?
Modul AGENTA zeigt dir, wie. Praxisnah & mit klaren technischen Leitplanken.
đEntdecken: https://software-architecture-camp.de/modul-agenta-architektur-fuer-agentic-software-engineering/
@wideareaai.bsky.social
@stsquad.mastodon.org.uk.ap.brid.gy#codingagents are so verbose. Almost everything they generate can be cut down to the bare essentials. The fact "they" are so defensive is basically them hedging because "they" don't truly understand the code they are editing.
The triage bot keeps resurrecting old project labels despite [âŚ]
@myfear.comI wired IBM Bob Shell v2 into GitHub Actions to turn Markdown into a validated GitHub Pages site: pinned TasteSkill, artifact deploy, no generated HTML in Git.
@cdevmkt.bsky.socialwhen an agent opens a PR, do you review like a human teammate or skim if CI is green.
shipping at chuongle.dev i read the weird files first.
curious what your bar is.
#codereview #codingagents #indiedev #devtools
@wideareaai.bsky.social@wideareaai.bsky.social@o-zehentleitner.bsky.socialYour repository already remembers almost everything about your project:
code, docs, tests, history, structure.
What it usually doesn't preserve is why.
AI didn't create that problem. It made capturing the missing layer cheap.
@alekseibalchunas.bsky.socialI love Codexâs remote feature when it works. The problem is, way too often it doesnât.
@tobosoksini.gamesDid a deep dive into why Blueprints are actually being deprecated in #UE6: tobosoksini.games/blueprints-w...
#UnrealEngine5 #UnrealEngine6 #GameDev #Blueprints #Verse #GameEngine #UE5 #AI #AIagents #CodingAgents