Bash worktree cwd workflow record

Status

Status: waiting
Current phase: 2. Discover and decide requirements
Next safe action: obtain approval for the Pi dependency upgrade scope and implementation plan before changing dependencies.

Request: address the first repository TODO so Bash execution honors the current Pi tool context inside hidden Git worktrees.
In scope: reproduce the Pi 0.84.4 behavior, adopt the upstream Pi 0.85.0 cwd correction through the repository dependency set, test workspace-root and nested-package execution in a disposable hidden worktree, update affected documentation, and remove the resolved TODO.
Out of scope: adding a custom Bash override, changing Bash's public per-call arguments, publishing packages, deploying, pushing, or modifying live/production data.

Source branch: main
Base commit: 7e7cfdb5cb0e1a1218b9a2c8250267d5099637fc
Feature branch: feature/honor-bash-worktree-cwd
Worktree: /srv/data/code/jb-pi/harness-tools/.worktrees/honor-bash-worktree-cwd
Omitted source changes: none; source checkout was clean and equal to origin/main after git fetch origin main.

The repository resolves a Pi Coding Agent version whose built-in Bash tool uses the execution ctx.cwd instead of the tool factory's original process directory.
A regression test executes pwd from both a disposable hidden worktree root and a nested workspace package directory and observes the requested directory.
A missing requested working directory fails before the shell command can create a sentinel side effect.
Existing repository typecheck and test suites pass.
The resolved TODO and temporary workaround language are removed or updated without changing unrelated historical evidence.

Plan/specification source: user request, docs/todos.md, and upstream Pi 0.84.4/0.85.0 source under /srv/data/opensrc/repos/github.com/earendil-works/pi/.
Exact plan: pending approval. Proposed: (1) add a focused repository regression harness for dynamic Bash ctx.cwd with hidden-worktree root, nested package, and missing-directory cases; (2) update the aligned @earendil-works/pi-ai, @earendil-works/pi-coding-agent, and @earendil-works/pi-tui catalog versions to 0.85.0 and regenerate bun.lock; (3) run focused tests, typecheck, and the full suite; (4) update README compatibility text and resolve the TODO/workaround references; (5) verify in a fresh Pi 0.85.0 process without changing the currently running global Pi installation.
Plan approval: pending.
UI classification and direction approval: non-UI; no prototype gate applies.
Risk decisions and scope: dependency upgrade requires explicit approval. Global Pi installation changes are excluded unless separately authorized.

Assignment	Owner identity	Deadline	Exclusive paths	Before/after Git evidence	Outcome
Workflow setup and research	parent Pi session `01a06e08-8d07-76a7-a116-d7e3baf55b87`	15 minutes	workflow record only	Before: clean `main` at `7e7cfdb`; after: pending initial record commit	In progress

Surface	Command	Exit code	Outcome
Source/cache preflight	`git status --short --branch`; `opensrc path ...`	0	Clean source; Pi 0.84.4 and 0.85.0 source available locally.
Focused cwd regression	pending	pending	pending
Typecheck	`bun run typecheck`	pending	pending
Full tests	`bun run test`	pending	pending
Fresh-process integration	pending	pending	pending

Synchronization and re-verification: fetch/rebase onto latest origin/main, then rerun affected and full verification if the effective tree changes.
Fast-forward result: pending
Product and closure commits: pending
Reachability and cleanup: pending
Rollback strategy: before integration, delete only the isolated feature branch/worktree after proof and approval if abandoned; after integration, revert the focused dependency/test/docs product commit.

Last committed valid checkpoint: none yet
Blockers: dependency-upgrade and exact-plan approval.
Exact next safe action: commit this initial record, present it, and ask the user to approve or reject the proposed dependency scope and plan.