# doer seed prompts — self-referential, tool-heavy, pipe-friendly
# one prompt per line; blank lines and lines starting with # are ignored

# === basic pipe / stdin patterns ===
translate this to turkish, one word
summarize in 3 bullet points
extract all email addresses
count tokens approximately
detect language and reply with ISO code only
format as JSON array
convert to uppercase
reverse word order

# === shell / devops ===
show current git branch and last 3 commits
list all Python files changed in the last week
find TODO comments in the project
run the test suite and summarize pass/fail
check which ports are open on localhost
diff the last two commits on main
who has pushed to main this month
identify the biggest files in this directory
tail the most recent log file

# === coding tasks ===
write a one-liner that counts unique lines in a file
fix the failing test described in the piped stacktrace
explain this error in one sentence
suggest a better variable name for `x`
refactor this function to be pure
write a pytest for the function in stdin
generate a regex that matches ISO dates
make this code idiomatic Python

# === self-awareness (doer-specific) ===
what's your file path? just the path
what model are you using right now
list your environment variables that start with DOER_
how many training turns do you have so far
what provider auto-detect chose and why
read your own SOUL.md and summarize it in one sentence
what's the current working directory
which AGENTS.md section controls tool calls

# === agentic / multi-step ===
read README.md and check if the install instructions work
find every env var documented in AGENTS.md and verify it's used in the code
list all functions in doer/__init__.py with their signatures
generate a changelog entry for the last 3 commits
write a commit message for the uncommitted diff
identify public API vs internal helpers in the main module
look for dead code in the doer package

# === training / dataset meta ===
estimate my training corpus size
show me the last 3 training records
what's in ~/.doer_history
how big is ~/.doer_training.jsonl
list the top 10 most common query patterns in training history
verify the training dataset has no null queries
count how many records have image attachments

# === minimalism / philosophy ===
in one sentence: what does doer do
why does doer refuse to add config files
explain: "context over memory"
why is the source code injected into every system prompt
give me the doer elevator pitch in 15 words

# === quick answers (style calibration) ===
ok
yes
no
done
wat
pwd in one word
python version
