Review the agent prompt templates in agentfox/_templates/profiles. There are instructions in the prompts that only make sense if the project that is being worked on is a python project.
Example (file coder_fix.md):
- Run the linter:
uv run ruff check <changed-files>
- Run tests to verify your fix. Prefer targeted subset runs over full suite
runs whenever possible (e.g., uv run pytest tests/unit/nightshift/)
Example (file agent.md):
## Quality Commands
| Command | What it does |
|---------|-------------|
| `make check` | Run lint + all tests (use before committing) |
| `make test` | Run all tests (`uv run pytest -q`) |
Expected change
Each specification defines a set of test and QA commands in tasks.json. Instead of "hard-coding" tools in the prompt files, change their language in such a way that the agent is instructed to use the appropriate tool as defined in tasks.json. This way the prompts are generic and clean and the agent gets correct, project/context aware instructions.
Review the agent prompt templates in
agentfox/_templates/profiles. There are instructions in the prompts that only make sense if the project that is being worked on is a python project.Example (file
coder_fix.md):uv run ruff check <changed-files>runs whenever possible (e.g.,
uv run pytest tests/unit/nightshift/)Example (file
agent.md):Expected change
Each specification defines a set of test and QA commands in
tasks.json. Instead of "hard-coding" tools in the prompt files, change their language in such a way that the agent is instructed to use the appropriate tool as defined in tasks.json. This way the prompts are generic and clean and the agent gets correct, project/context aware instructions.