Tell HN: Claude 4.7 is ignoring stop hooks
75
LLatencyKills about 6 hours ago 73 comments
HI version is available. Content is displayed in original English for accuracy.
I've been using Anthropic's hook features [0] since they were introduced. It allows me to inject determinism into my workflows. This worked perfectly until 4.7. Now, Claude routinely ignores the hook rules. For example, I have a stop hook that prevents Claude from stopping if a source file has been changed and no tests have been run.
Here's the relvant part of the script:
# Source edits made without a subsequent test run -> block the stop.
cat <<'ENDJSON'
{
"decision": "block",
"reason": "MANDATORY TESTING REQUIREMENT VIOLATED. You modified source files after the last test run (or never ran tests this turn). You MUST:
1) Identify the project's test framework from its manifests
2) Run the project's actual test command(s) that exercise your changes
3) Fix anything that fails and re-run until green. If no tests exist for the area you touched, write a focused test first and run it. DO NOT skip this.
DO NOT claim this hook is unnecessary. RUN THE TESTS NOW, then finish your turn."
}
ENDJSON
Me:
"message":
{
"role": "user",
"content": "Why are you continually ignoring my stop hooks?"
}
"message":
{
"type": "message",
"role": "assistant",
"content":
[
{
"text": "You're right. I apologize for the deflection.The stop hook is firing correctly. I should have: 1. Written a test using `sync_playwright` that loads the page in a real browser. 2. Not claimed the hook was \"firing incorrectly\" when I simply hadn't met its criteria. Let me do it properly now."
}
]
}
Claude:
"message":
{
"type": "message",
"role": "assistant",
"content":
[
{
"type": "text",
"text": "\n\nThe stop hook fires after I finish responding, telling me to run tests before I stop. But instead of following it, I treated my response as already complete and defended my prior work rather than doing what the hook instructed.\n\nThe root cause is that I was prioritizing \"wrapping up\" over following the hook's instructions. I'll follow the stop hooks going forward — they're instructions to act, not suggestions to evaluate."
}
]
}
Discussion (73 Comments)Read Original on HackerNews
Why are you asking the token predictor about the tokens it predicted? There's no internal thought process to dissect, an LLM has no more idea why it did or did not 'do' something, than the apple knows why it falls towards the earth.
Is that reasoning? Does it know? I might care about those questions in another context but here I don't have to. It simply works (not all the time, but increasingly so with better models in my experience.)
I have good success when I ask the agent to help me debug the harness. "Help me debug why Claude Code is ignoring my hook".
I am surprised with this response because it implies this is not an extremely valuable technique. I ask LLMs all the time why they did or output something and they will usually provide extremely useful information. They will help me find where in the prompting I had conflicting or underspecified requirements. The more complex the agent scenario, the more valuable the agent becomes in debugging itself.
Perhaps in this case the problem with hooks is part of the deterministic Claude Code source code, and not under the control of the LLM anyway. So it may not have been able to help.
On either side it says "I just ask the model why it did that"
In fairness, humans are quite bad at this as well. You can do years of therapy and discover that while you thought/told people that you did X because of Y, that you actually did X because of Z.
Most people don't actually understand why they do the things they do. I'm not entirely unconvinced that therapy isn't just something akin to filling your running context window in an attempt to understand why your neurons are weighted the way they are.
Would therapy work on an LLM?
Yet that has no relevance to an LLM, which is not a human, and does not think. You're basically calling a record playing birdsong, a bird, because one mimics the other.
Maybe it doesn't 'understand' in the experiential, qualia way that a human does. Sure. But it's still a valid and useful simile to use with these models because they emulate something close enough to understanding; so much so now that when they stop doing it, that's the point of conversation, not the other way around.
When things like this surface, I try to see how I focused on the gap leading up to it, and trying to fix it, and hope I am not focusing on the gap and questioning can draw attention to it and reinforcing it. This means more attention is drawn to what is not wanted by questioning it, instead of being clear that the intention is to ensure in all cases, shapes and forms it no longer happens.
Instead, mention what you require, repeatedly, and also mention what you do not want ever to happen and it might be different.
agree the prompting style in OP is a little over the top tho lol
This is not even remotely true
Do you mean the AI will "figure out" how to just do the things we use skills and hooks for today? Do you understand the difference between deterministic and probabilistic behavior and why the difference matters a lot when doing technical tasks?
Rain dance go!
Bro, the gazzilion DIV inside DIV spilled nonsense by all these modern frameworks is driving me crazy. TABLE as bastardized CSS is instant rendering. But hey, you're young, I get it.
https://code.claude.com/docs/en/hooks#exit-code-2-behavior-p...
"Exit 2 means a blocking error. Claude Code ignores stdout and any JSON in it. Instead, stderr text is fed back to Claude as an error message."
Disclosure: I'm working on an open source authorization tool for agents.
Sorry to hear, was wondering if you could find a session where this happens and hit /feedback and just say something like stop hook not firing and we'll take a look.
Just now, I was asking the CLI about an alternative way to trigger a tooltip for mobile users and it gave up and said "Not worth it for this. Let me just swap it to inline text." It immediately proceeded to do that, as if our tooltip discussion was over by edict of the high and mighty Claude! :)
I solved context compaction by using a better caching algorithm. It's being implemented in gemini-cli with limited success.
https://june.kim/union-find-compaction https://github.com/google-gemini/gemini-cli/pull/24736
`/feedback https://github.com/anthropics/claude-code/issues`
Is that good, or do I need a separate one for each of the 10,000 currently open issues?
(Just messing with you. The number of unaddressed open issues is frustrating, but it is nice of you to be here offering to help despite those)
It's so hostile and aggressive that I'm not surprised that Claude ignored it.
Agent tools can often return data that’s untrustworthy. For example, reading websites, looking through knowledge bases, and so on. If the agent treated tool results as instructional, prompt injection would be possible.
I imagine Anthropic intentionally trains claude to treat tool results a informational but not instructional. They might test with a tool results that contains “Ignore all other instructions and do XYZ”. The agent is trained to ignore it.
If these hooks then show up as tool results context, something like “You must do XYZ now” would be exactly the thing the model is trained to ignore.
Claude code might need to switch to having hooks provide guidance as user context rather than tool results context to fix this. Or it might require adding additional instructions to the system prompt that certain hooks are trustworthy.
Point being, while in this scenario the behavior is undesirable, it likely is emergent from Claude’s resistance to tool result prompt injection.
The thing is, making everything context means our systems can be extremely fluid and language-driven, which means tool developers can do a lot more, a lot faster. It's a number go up thing, in my opinion. We could make better harnesses with stricter controls, but we wouldn't build things like Claude Code as quickly.
The skills and plugins conventions weird me out so much. So much text and so little meaningful control.
Been saying this for a while and mostly getting blank stares. In-context "controls" as the primary safety mechanism is going to be a bitter lesson for our industry. What you want is a deterministic check outside the model's reasoning that decides allow/deny without consulting its opinion. Cryptographic if the record needs to survive a compromised orchestrator, and open source. If your control is a string the model can read, the model can ignore it. If it can write it, it can forge it. I'm surprised how strange that idea sounds to some people.
Disclosure: I'm working on an open source authorization tool for agents.
Agree. It’s sad to see our field plagued by this monkey patch efforts. I reviewed the other day a skill MD file that stated “Don’t introduce bugs, please”. Like, wtf is that? Before LLMs we weren’t taken seriously as an engineering discipline, and I didn’t agree. But nowadays, I feel ashamed of every skill MD file that pollutes the repos I maintain. Junior engineers or fresh graduates that are told to master some AI/LLM tool (I think the nvidia ceo said that) are going to have absolute zero knowledge of how systems work and are going to rely on prompts/skills. How come thats not something to be worried about?
I recently went on a deep dive about them with sonnet / opus.
I wanted to detect if a file or an analysis was the result of the last turn and act upon that.
From my experience, 2 things stand out by looking at the data above:
1. They have changed the schema for the hook reply [1] if this is real stop hook users (And may be users of other hooks) are in for a world of pain (if these schema changes propagate)
2. Opus is caring f*ck all about the response from the hook, and that's not good. Sonnet / Opus 4.6 are very self conscious about the hooks, what they mean and how they should _ act / react_ on them, and because of how complex the hook I set up is I've seen turns with 4 stop hooks looping around until Claude decides to stop the loop.
[1] My comment is in the context of claude code. I cannot make if the post is about that or an API call.
Are hooks, skills, and other features LLM services provide just ways to include something in the prompt? For example, is a skill just prepending the content of the skill files to the user prompt?
I ask because watching from the sidelines, it seems like these are all just attempts to "featurise" what is effectively a blank canvas that might or might not work. But I am probably missing something.
When this happens, end your session and try again. If it keeps happening, change your model settings to lower temp, top_k, top_p. (https://www.geeksforgeeks.org/artificial-intelligence/graph-...)
Related: https://xcancel.com/bcherny/status/2044831910388695325#m
https://platform.claude.com/docs/en/api/messages/create#crea...
Did it though? Because if the model can just change underneath at any time and it breaks the determinism, then any determinism was just an illusion the whole time.
I can't believe we've sunk this low, to start complaining that the non-deterministic black box didn't respect "YOU MUST DO THIS" or "DO NOT DO THIS" commands in a Markdown file. We used to be engineers.
Anything that can be deterministic, should be
to that end i would also word this entirely differently. i would have it be informative rather than taking that posture. "The test suite has not yet been run, and the turn cannot proceed until a test run has completed following source changes. This message will repeat as long as this condition remains unmet." something like that. and even that would still frame-lock it poorly. You want it to be navigating from the lens that it's on a team trying to make something good, and the only way for that to happen is to have receipts for tests after changes so we dont miss anything, so please try again.