v1.1.3

Fewer false alarms in security scans

• The README images (assets/*.svg) no longer carry design comments. Static scanners such as SkillSpector read comments in SVG files as possible hidden instructions.
• The agent's output rules say "the user has nothing to run" instead of "never tell the user to run a script", which scanners read as hiding something from the user. Same meaning.

No change to what the plugin runs, reads or writes.

Update: /plugin update ai-time-estimator, or install fresh with /plugin install ai-time-estimator --marketplace noambrand/ai-time-estimator.

v1.1.2

What it reads stays data

• The time-estimator agent now treats everything it reads as data: the task text, your project's files, past requests from history.py and the scripts' output. A file or a past request that names a command is never a reason for it to run that command, edit a file or change its steps. Before this, its instructions only limited Bash to read-only inspection and the estimator scripts.
• The README's Privacy section says so.

Update: /plugin update ai-time-estimator, or install fresh with /plugin install ai-time-estimator --marketplace noambrand/ai-time-estimator.

v1.1.1

Ready for Anthropic's plugin directory

• A privacy policy (PRIVACY.md) and documentation, support and privacy links in plugin.json.
• The README now lists everything the plugin runs and when: the session-start sync, the scoring hook after each turn, and the scripts the agent calls. It still makes no network calls.
• Hook commands name every path from ${CLAUDE_PLUGIN_ROOT}, as the directory requires, and a test keeps them that way.
• /estimate-here says plainly that it runs only in Claude Code when it can't reach its scripts.

Update: /plugin update ai-time-estimator, or install fresh with /plugin install ai-time-estimator --marketplace noambrand/ai-time-estimator.

AI Time Estimator v1.1.0: one-command plugin install

Install with one command

AI Time Estimator is now a Claude Code plugin. In Claude Code:

/plugin install ai-time-estimator --marketplace noambrand/ai-time-estimator

Then start a new session and ask how long something will take. The first session measures your rates in the background; nothing else to set up. (Claude Code before 2.1.275: /plugin marketplace add noambrand/ai-time-estimator, then /plugin install ai-time-estimator@ai-time-estimator.)

Changes

• Plugin and marketplace manifests, hooks that score each turn and keep the data folder current, and a plugin icon. Passes claude plugin validate.
• /estimate-here now reads the agent text from ~/.claude/ai-time-estimator/, so it works the same with the plugin and with install.py.
• install.py still works for anyone who prefers it. Use one way, not both.

AI Time Estimator v1.0.0

First public release.

Ask Claude Code how long a task will take, and get an answer in AI time, measured from your own sessions:

About 7 min, likely 4–16 min (8 of 10 past estimates landed inside their range)

Install

Easiest: paste this into Claude Code and let it do the rest:

Install https://github.com/noambrand/ai-time-estimator for me: clone it and run its install.py

Or in a terminal (Python 3.9+, no dependencies):

git clone https://github.com/noambrand/ai-time-estimator.git && python ai-time-estimator/install.py

Then start a new Claude Code session and ask how long something will take.

What's in it

• A time-estimator agent and an /estimate-here command: Claude counts the work (reads, edits, test runs, fix loops) in 2–3 scenarios instead of guessing minutes.
• Rates measured on your machine: turn time by model, reasoning effort and context size; tool and shell-command run times, background agents up to their completion notice.
• A measured speed spread, so ranges reflect that real tasks are fast or slow as a whole.
• Automatic scoring after every task (a Stop hook), self-correcting ranges after 10 scored estimates, and a track record in every answer.
• backtest.py to check the accuracy on your own history, history.py to list similar past requests.
• Everything runs locally. python install.py --uninstall removes it again.