feat(workflows): experimental GSD workflow suite (gsd-*) — port from open-gsd/gsd-core#1986
Open
uinstinct wants to merge 3 commits into
Open
feat(workflows): experimental GSD workflow suite (gsd-*) — port from open-gsd/gsd-core#1986uinstinct wants to merge 3 commits into
uinstinct wants to merge 3 commits into
Conversation
…open-gsd/gsd-core Closes coleam00#1696 11 workflows porting the GSD core loop: map-codebase, new-project, discuss/plan/execute/verify phases, ship, complete-milestone, new-milestone, plus quick and debug. 10 shared role prompts in .archon/commands/gsd/ for agent context. Placed flat as .archon/workflows/experimental/gsd-*.yaml (not in a subdirectory) because workflow discovery deliberately caps at depth 1 (workflow-discovery.ts:89). All workflows run with worktree.enabled: false — .planning/ state + built code must persist across the execute→verify→ship runs. .planning/.gitignore excludes GSD scratch files.
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (12)
✅ Files skipped from review due to trivial changes (1)
🚧 Files skipped from review as they are similar to previous changes (4)
📝 WalkthroughWalkthroughPorts the GSD methodology into Archon: adds role command specs, interactive workflows for map→discuss→plan→execute→verify→ship, supporting utilities (debug/quick/milestone), and .planning state artifacts with approval gates and commit protocols. ChangesExperimental GSD Workflow Suite
Estimated code review effort🎯 4 (Complex) | ⏱️ ~75 minutes Possibly related PRs
Poem
✨ Finishing Touches🧪 Generate unit tests (beta)
|
Author
|
@Wirasm If we can merge this it will be awesome. GSD is a core workflow to many projects |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 7
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (2)
.archon/workflows/experimental/gsd-execute-phase.yaml (1)
261-341:⚠️ Potential issue | 🟠 Major | ⚡ Quick winSubshell variable assignment won't persist —
PHASE_DIRwill always be empty.Lines 271-273 attempt to set
PHASE_DIRinside a while loop in a pipeline:PHASE_DIR=$(ls -d .planning/phases/*/ 2>/dev/null | sort -V | while read d; do if [ -f "${d}VERIFICATION.md" ]; then PHASE_DIR="$d"; fi done)The
whileloop runs in a subshell (due to the pipe), so thePHASE_DIR="$d"assignment doesn't persist. Additionally, the command substitution captures stdout, but nothing is printed — the assignment is silent. Result:PHASE_DIRis always empty.Suggested fix
if [ -z "$PHASE_NUM" ]; then # Best-effort: find the highest-numbered phase dir with a VERIFICATION.md - PHASE_DIR=$(ls -d .planning/phases/*/ 2>/dev/null | sort -V | while read d; do - if [ -f "${d}VERIFICATION.md" ]; then PHASE_DIR="$d"; fi - done) - if [ -n "$PHASE_DIR" ]; then + PHASE_DIR="" + for d in $(ls -d .planning/phases/*/ 2>/dev/null | sort -V); do + if [ -f "${d}VERIFICATION.md" ]; then + PHASE_DIR="$d" + fi + done + if [ -n "$PHASE_DIR" ]; then PHASE_NUM=$(basename "$PHASE_DIR" | grep -oE '^[0-9]+') fi🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.archon/workflows/experimental/gsd-execute-phase.yaml around lines 261 - 341, The PHASE_DIR assignment inside the pipeline/subshell (the "ls -d ... | sort -V | while read d; do ... PHASE_DIR=\"$d\"; done") won't persist because the while runs in a subshell and nothing is printed for command substitution; replace that pipeline with a non-subshell iteration (e.g., a for loop or while reading via process substitution) that directly assigns PHASE_DIR in the current shell, or write the matched directory to stdout for capture, so the block that sets PHASE_DIR when ARGUMENTS/PHASE_NUM is empty correctly produces a value..archon/workflows/experimental/gsd-verify-work.yaml (1)
286-315:⚠️ Potential issue | 🟠 Major | ⚡ Quick winBash node cannot access previous node outputs via dot notation.
Lines 289-293 attempt to reference guard node outputs in a bash script:
DIR=$guard.output.PHASE_DIR PHASE_BASE=$guard.output.PHASE_BASE ... UAT_FILE=$guard.output.UAT_FILEIn bash,
$guardis an undefined variable (empty), so these lines evaluate to:
DIR=.output.PHASE_DIR(literal string)PHASE_BASE=.output.PHASE_BASE- etc.
The route node needs to re-derive these values from the filesystem or receive them through a different mechanism (e.g., environment variables set by the workflow engine).
Suggested fix
- id: route depends_on: [uat] bash: | - DIR=$guard.output.PHASE_DIR - PHASE_BASE=$guard.output.PHASE_BASE - - # Count failures from UAT summary - UAT_FILE=$guard.output.UAT_FILE + # Re-derive phase directory (same logic as guard node) + if [ -n "$ARGUMENTS" ]; then + PHASE_NUM="$ARGUMENTS" + case ${`#PHASE_NUM`} in + 1) PAD="0${PHASE_NUM}" ;; + *) PAD="${PHASE_NUM}" ;; + esac + DIR=$(ls -d .planning/phases/${PAD}-* 2>/dev/null | head -1) + [ -z "$DIR" ] && DIR=$(ls -d .planning/phases/${PHASE_NUM}-* 2>/dev/null | head -1) + else + DIR=$(find .planning/phases/ -maxdepth 2 -name "*-SUMMARY.md" 2>/dev/null | sort -t/ -k3,3 | tail -1 | xargs dirname) + fi + + PHASE_BASE=$(basename "$DIR") + UAT_FILE="$DIR/${PHASE_BASE}-UAT.md" + + # Count failures from UAT summary FAILED=$(grep -c '| FAIL ' "$UAT_FILE" 2>/dev/null || echo 0)🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.archon/workflows/experimental/gsd-verify-work.yaml around lines 286 - 315, The bash step in the route node is incorrectly trying to read previous node outputs via "$guard.output.PHASE_DIR", "$guard.output.PHASE_BASE" and "$guard.output.UAT_FILE" (these are not valid shell variables); update the route node to obtain those values correctly by either reading the produced files from the PHASE_DIR on disk or accepting them via environment variables passed into the bash step by the workflow engine — replace uses of DIR, PHASE_BASE and UAT_FILE to derive values from the filesystem (e.g., find or fixed paths under the workspace/PHASE_DIR) or from explicitly exported env vars (e.g., PHASE_DIR, PHASE_BASE, UAT_FILE) so the script references real shell variables like $PHASE_DIR, $PHASE_BASE and $UAT_FILE instead of $guard.output.*.
🧹 Nitpick comments (2)
.archon/commands/gsd/gsd-verifier.md (1)
93-99: 💤 Low valueBash
basenamesyntax issue in documentation example.The
basenamecommand only accepts a single suffix argument. The pattern$(basename "$artifact" .tsx? .ts)won't work as intended —basenamewill treat.tsx?as the suffix and ignore.ts.Since this is specification documentation showing conceptual verification commands, consider clarifying the intent or fixing the example:
Suggested clarification
**Level 3 — Wiring:** ```bash # Imported? -grep -r "import.*$(basename "$artifact" .tsx? .ts)" src/ --include="*.ts" --include="*.tsx" 2>/dev/null +# Strip extension to get component name (handle .ts, .tsx, .js, .jsx) +COMPONENT=$(basename "$artifact" | sed 's/\.\(tsx\?\|jsx\?\|ts\|js\)$//') +grep -r "import.*$COMPONENT" src/ --include="*.ts" --include="*.tsx" 2>/dev/null # Used beyond imports? -grep -r "$(basename "$artifact" .tsx? .ts)" src/ --include="*.ts" --include="*.tsx" 2>/dev/null | grep -v "import" +grep -r "$COMPONENT" src/ --include="*.ts" --include="*.tsx" 2>/dev/null | grep -v "import"🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.archon/commands/gsd/gsd-verifier.md around lines 93 - 99, The example uses incorrect basename usage: replace the single-call basename pattern `$(basename "$artifact" .tsx? .ts)` with a small shell step that strips extensions into a variable (e.g., compute COMPONENT from basename and remove .ts/.tsx/.js/.jsx via a sed or parameter expansion), then reuse that COMPONENT in the two grep commands; specifically, create COMPONENT from basename("$artifact") with extension removal (or use `${artifact%.*}`/sed), then use grep -r "import.*$COMPONENT" ... and grep -r "$COMPONENT" ... | grep -v "import" so both the import check and usage check reference the same cleaned component name instead of the invalid basename call..archon/commands/gsd/gsd-roadmapper.md (1)
51-55: 💤 Low valueConsider adding language identifiers to fenced code blocks.
The code blocks at lines 51-55 and 105-111 are missing language identifiers. While these are conceptual examples rather than executable code, adding a language identifier (e.g.,
textormarkdown) improves consistency and helps syntax highlighters.Suggested fix
-``` +```text Phase 1: All database models ← Too coupled-``` +```text AUTH-01 → Phase 2Also applies to: 105-111
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.archon/commands/gsd/gsd-roadmapper.md around lines 51 - 55, Add language identifiers to the fenced code blocks that currently show conceptual lines like "Phase 1: All database models ← Too coupled" and "AUTH-01 → Phase 2" by replacing the plain backtick fences with fences that include a language (e.g., use ```text or ```markdown) so syntax highlighters render them consistently; update both occurrences (the block containing the "Phase 1..." text and the block containing "AUTH-01 → Phase 2") to include the chosen language identifier.Source: Linters/SAST tools
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In @.archon/workflows/experimental/gsd-complete-milestone.yaml:
- Around line 346-375: The workflow currently instructs creation at
`.planning/MILESTONES.md` but stages `MILESTONES.md` at repo root, causing an
uncommitted file; pick one canonical path (prefer the repo-root `MILESTONES.md`
to match staging behavior), update the "Step 7: Update MILESTONES.md"
instructions and any template/header text to reference `MILESTONES.md` (not
`.planning/MILESTONES.md`), and ensure subsequent staging/commit steps and any
other occurrences (the later block around the staging list currently adding
`MILESTONES.md`) consistently use that same `MILESTONES.md` path.
In @.archon/workflows/experimental/gsd-debug.yaml:
- Around line 73-79: The slug derivation (SLUG) can collide or become empty due
to truncation and non-alphanumeric collapse; change the generation to append a
stable short fingerprint so filenames are unique and still human-readable: keep
the sanitized base (from ARGUMENTS) but if it is empty use a safe fallback
token, then append a short hash or short timestamp+random suffix (e.g., first 8
chars of sha256(ARGUMENTS) or epoch+rand) to form the final SLUG, and update
references to .planning/debug/$SLUG.md; ensure the sanitization step still
collapses non-alnum to hyphens and the final result is truncated only after the
hash is appended so collisions are avoided.
In @.archon/workflows/experimental/gsd-execute-phase.yaml:
- Around line 79-91: The shell glob is missing a directory separator between the
directory variable and the pattern: update the glob usages that reference
PHASE_DIR and d so they match files inside the directory (e.g., ensure the
pattern inserts a "/" between PHASE_DIR and *-PLAN.md and likewise between d and
*-PLAN.md). Fix all ls calls and the ls used in the Plan Inventory loop that use
"$PHASE_DIR"*-PLAN.md and the earlier "$d"*-PLAN.md so they correctly expand to
files inside the target directory rather than concatenating the name and the
pattern.
In @.archon/workflows/experimental/gsd-new-milestone.yaml:
- Around line 364-375: Don’t derive the next phase from the working tree
(.planning/phases) in this block (PHASE_DIR / PHASE_NUM); instead determine
NEXT_PHASE from the archived milestone data (e.g., use an existing archived
milestone variable or compute from the archived milestone’s last phase) and only
fall back to 1 if no archived milestone info exists; update the echo lines and
the suggested command to use gsd-discuss-phase $NEXT_PHASE so the workflow
continues numbering from the archived milestone rather than assuming phase 1
when no local phase dirs exist.
In @.archon/workflows/experimental/gsd-new-project.yaml:
- Around line 405-444: The roadmap node in the workflow conflicts with the
gsd-roadmapper.md role file which instructs the roadmapper to commit; update
gsd-roadmapper.md (the Commit section) to make committing conditional or remove
the automatic commit so it honors the workflow’s approval gate: either change
the Commit steps into a conditional (e.g., only commit when an explicit flag/ENV
like COMMIT_AFTER_APPROVAL is set) or replace the commit commands with
instructions to stage/save files for later (e.g., “write files but do not
commit; finalize node will commit”), and document the new expected behavior in
gsd-roadmapper.md so the roadmap agent (roadmapper role) and the finalize node
remain consistent.
In @.archon/workflows/experimental/gsd-plan-phase.yaml:
- Around line 238-251: The sed is being run against the full "Next: ..." string
so it never matches; instead extract the phase basename from PHASE_DIR and
derive the numeric prefix. Replace the pipeline that echoes "Next: ..." with
logic that captures the basename (using ${PHASE_DIR##*/}) into a variable (e.g.,
BASENAME) and then either use parameter expansion to get the phase number
(PHASE_NUM=${BASENAME%%-*}) or run sed against just the basename (echo
"${BASENAME}" | sed 's/^\([0-9][0-9]\).*/\1/'); then print the "Next: archon
workflow run gsd-execute-phase $PHASE_NUM" message. Ensure you update the line
that currently uses echo "Next: ... ${PHASE_DIR##*/}" | sed 's/^\(..\)-.*/\1/'
to the corrected extraction and echo.
In @.archon/workflows/experimental/gsd-ship.yaml:
- Line 82: The VSTATUS assignment uses GNU-only grep with -P; replace the
command that sets VSTATUS (currently using grep -oP 'status:\s*\K\S+' "$VFILE")
with a portable alternative (for example an awk or sed pipeline) that extracts
the first non-space token after "status:" from VFILE and preserves the existing
fallback (|| echo "unknown"); update the VSTATUS assignment to use that portable
extraction and still redirect errors to /dev/null as before.
---
Outside diff comments:
In @.archon/workflows/experimental/gsd-execute-phase.yaml:
- Around line 261-341: The PHASE_DIR assignment inside the pipeline/subshell
(the "ls -d ... | sort -V | while read d; do ... PHASE_DIR=\"$d\"; done") won't
persist because the while runs in a subshell and nothing is printed for command
substitution; replace that pipeline with a non-subshell iteration (e.g., a for
loop or while reading via process substitution) that directly assigns PHASE_DIR
in the current shell, or write the matched directory to stdout for capture, so
the block that sets PHASE_DIR when ARGUMENTS/PHASE_NUM is empty correctly
produces a value.
In @.archon/workflows/experimental/gsd-verify-work.yaml:
- Around line 286-315: The bash step in the route node is incorrectly trying to
read previous node outputs via "$guard.output.PHASE_DIR",
"$guard.output.PHASE_BASE" and "$guard.output.UAT_FILE" (these are not valid
shell variables); update the route node to obtain those values correctly by
either reading the produced files from the PHASE_DIR on disk or accepting them
via environment variables passed into the bash step by the workflow engine —
replace uses of DIR, PHASE_BASE and UAT_FILE to derive values from the
filesystem (e.g., find or fixed paths under the workspace/PHASE_DIR) or from
explicitly exported env vars (e.g., PHASE_DIR, PHASE_BASE, UAT_FILE) so the
script references real shell variables like $PHASE_DIR, $PHASE_BASE and
$UAT_FILE instead of $guard.output.*.
---
Nitpick comments:
In @.archon/commands/gsd/gsd-roadmapper.md:
- Around line 51-55: Add language identifiers to the fenced code blocks that
currently show conceptual lines like "Phase 1: All database models ← Too
coupled" and "AUTH-01 → Phase 2" by replacing the plain backtick fences with
fences that include a language (e.g., use ```text or ```markdown) so syntax
highlighters render them consistently; update both occurrences (the block
containing the "Phase 1..." text and the block containing "AUTH-01 → Phase 2")
to include the chosen language identifier.
In @.archon/commands/gsd/gsd-verifier.md:
- Around line 93-99: The example uses incorrect basename usage: replace the
single-call basename pattern `$(basename "$artifact" .tsx? .ts)` with a small
shell step that strips extensions into a variable (e.g., compute COMPONENT from
basename and remove .ts/.tsx/.js/.jsx via a sed or parameter expansion), then
reuse that COMPONENT in the two grep commands; specifically, create COMPONENT
from basename("$artifact") with extension removal (or use `${artifact%.*}`/sed),
then use grep -r "import.*$COMPONENT" ... and grep -r "$COMPONENT" ... | grep -v
"import" so both the import check and usage check reference the same cleaned
component name instead of the invalid basename call.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 39a9f52b-f9ee-475c-94fe-4c704d27a561
📒 Files selected for processing (23)
.archon/commands/gsd/gsd-codebase-mapper.md.archon/commands/gsd/gsd-debugger.md.archon/commands/gsd/gsd-executor.md.archon/commands/gsd/gsd-phase-researcher.md.archon/commands/gsd/gsd-plan-checker.md.archon/commands/gsd/gsd-planner.md.archon/commands/gsd/gsd-project-researcher.md.archon/commands/gsd/gsd-research-synthesizer.md.archon/commands/gsd/gsd-roadmapper.md.archon/commands/gsd/gsd-verifier.md.archon/workflows/experimental/gsd-complete-milestone.yaml.archon/workflows/experimental/gsd-debug.yaml.archon/workflows/experimental/gsd-discuss-phase.yaml.archon/workflows/experimental/gsd-execute-phase.yaml.archon/workflows/experimental/gsd-map-codebase.yaml.archon/workflows/experimental/gsd-new-milestone.yaml.archon/workflows/experimental/gsd-new-project.yaml.archon/workflows/experimental/gsd-plan-phase.yaml.archon/workflows/experimental/gsd-quick.yaml.archon/workflows/experimental/gsd-ship.yaml.archon/workflows/experimental/gsd-verify-work.yaml.planning/.gitignoreCHANGELOG.md
Shell-script correctness fixes in the gsd-* workflow bash nodes and role docs:
- gsd-execute-phase: add the missing `/` in the PLAN globs so they match
files inside the phase dir (not siblings); replace the lost-subshell
PHASE_DIR assignment in the report node with a current-shell for-loop.
- gsd-verify-work: the guard now emits a single JSON object on stdout
(diagnostics → stderr) so the engine resolves $guard.output.FIELD in the
uat prompt and route node. A schemaless bash producer with non-JSON output
throws OutputRefError on .FIELD access, which previously failed uat.
- gsd-plan-phase: derive the phase number via basename + grep instead of a
sed pattern that never matched the "Next: ..." string.
- gsd-ship: replace GNU-only `grep -oP \K` with portable awk plus an explicit
${VSTATUS:-unknown} default.
- gsd-debug: collision-safe slug (<=21-char base + 8-hex git fingerprint) with
an empty-input guard for punctuation-only descriptions.
- gsd-complete-milestone: canonical repo-root MILESTONES.md (matches staging).
- gsd-new-milestone: derive the first phase from the freshly written ROADMAP.md
rather than the highest existing phase dir.
- gsd-roadmapper.md: stop committing (callers commit after their approval gate);
add a `text` language to the two bare code fences.
- gsd-verifier.md: fix the basename extension-stripping example (single sed -E).
All 11 gsd-* workflows remain schema-valid and prettier-clean; bash nodes
verified against fixtures.
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
.archon/workflows/experimental/gsd-plan-phase.yaml (1)
242-242:⚠️ Potential issue | 🟡 Minor | ⚡ Quick winGNU-only
find -printfmay break on macOS/BSD.The
-printf '%h\n'option is a GNU extension not available in BSD/macOSfind. Consider using a POSIX-portable alternative:- PHASE_DIR=$(find .planning/phases -name '*-PLAN.md' -printf '%h\n' 2>/dev/null | sort -u | tail -1) + PHASE_DIR=$(find .planning/phases -name '*-PLAN.md' 2>/dev/null | xargs -I{} dirname {} | sort -u | tail -1)Or using a shell loop:
PHASE_DIR=$(find .planning/phases -name '*-PLAN.md' 2>/dev/null | while read f; do dirname "$f"; done | sort -u | tail -1)🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.archon/workflows/experimental/gsd-plan-phase.yaml at line 242, The command that sets PHASE_DIR uses GNU-only find -printf '%h\n' which fails on BSD/macOS; update the PHASE_DIR assignment to use a POSIX-portable sequence instead (call find without -printf and pipe the results through dirname or a while-read loop, then sort -u and tail -1) so PHASE_DIR is derived from the parent directory of the latest '*-PLAN.md' in .planning/phases; edit the line that defines PHASE_DIR to replace the -printf usage with the portable pipeline (e.g., find ... | xargs -n1 dirname | sort -u | tail -1 or find ... | while read f; do dirname "$f"; done | sort -u | tail -1) ensuring error redirection (2>/dev/null) is preserved.
🧹 Nitpick comments (1)
.archon/workflows/experimental/gsd-debug.yaml (1)
261-262: 💤 Low valueArchive
git mvsource path has redundantmkdir -p.Line 261 runs
mkdir -p .planning/debugbut the session file is already at.planning/debug/{slug}.md(created by the setup node and used throughout). The mkdir is harmless but unnecessary. More importantly, verify the paths are consistent:
- Line 261:
mkdir -p .planning/debug && git mv .planning/debug/{slug}.md .planning/debug/resolved/The
mkdir -p .planning/debugshould bemkdir -p .planning/debug/resolvedif the intent is to ensure the destination exists (though setup already creates it at line 64).🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.archon/workflows/experimental/gsd-debug.yaml around lines 261 - 262, The git mv command is creating a resolved/ destination but the mkdir used before it is for .planning/debug (mkdir -p .planning/debug && git mv .planning/debug/{slug}.md .planning/debug/resolved/); update this to ensure the destination exists by changing the mkdir to mkdir -p .planning/debug/resolved or remove the mkdir entirely if the setup step already creates .planning/debug/resolved; confirm the command around the git mv that moves .planning/debug/{slug}.md to .planning/debug/resolved/ uses the same paths consistently.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Outside diff comments:
In @.archon/workflows/experimental/gsd-plan-phase.yaml:
- Line 242: The command that sets PHASE_DIR uses GNU-only find -printf '%h\n'
which fails on BSD/macOS; update the PHASE_DIR assignment to use a
POSIX-portable sequence instead (call find without -printf and pipe the results
through dirname or a while-read loop, then sort -u and tail -1) so PHASE_DIR is
derived from the parent directory of the latest '*-PLAN.md' in .planning/phases;
edit the line that defines PHASE_DIR to replace the -printf usage with the
portable pipeline (e.g., find ... | xargs -n1 dirname | sort -u | tail -1 or
find ... | while read f; do dirname "$f"; done | sort -u | tail -1) ensuring
error redirection (2>/dev/null) is preserved.
---
Nitpick comments:
In @.archon/workflows/experimental/gsd-debug.yaml:
- Around line 261-262: The git mv command is creating a resolved/ destination
but the mkdir used before it is for .planning/debug (mkdir -p .planning/debug &&
git mv .planning/debug/{slug}.md .planning/debug/resolved/); update this to
ensure the destination exists by changing the mkdir to mkdir -p
.planning/debug/resolved or remove the mkdir entirely if the setup step already
creates .planning/debug/resolved; confirm the command around the git mv that
moves .planning/debug/{slug}.md to .planning/debug/resolved/ uses the same paths
consistently.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: f2b79376-20d7-47a2-ae60-27f65fa2bdc5
📒 Files selected for processing (9)
.archon/commands/gsd/gsd-roadmapper.md.archon/commands/gsd/gsd-verifier.md.archon/workflows/experimental/gsd-complete-milestone.yaml.archon/workflows/experimental/gsd-debug.yaml.archon/workflows/experimental/gsd-execute-phase.yaml.archon/workflows/experimental/gsd-new-milestone.yaml.archon/workflows/experimental/gsd-plan-phase.yaml.archon/workflows/experimental/gsd-ship.yaml.archon/workflows/experimental/gsd-verify-work.yaml
✅ Files skipped from review due to trivial changes (1)
- .archon/commands/gsd/gsd-verifier.md
🚧 Files skipped from review as they are similar to previous changes (5)
- .archon/workflows/experimental/gsd-ship.yaml
- .archon/commands/gsd/gsd-roadmapper.md
- .archon/workflows/experimental/gsd-new-milestone.yaml
- .archon/workflows/experimental/gsd-complete-milestone.yaml
- .archon/workflows/experimental/gsd-execute-phase.yaml
Every work-producing experimental GSD workflow now ensures a non-base branch, pushes it, and creates or updates a pull request: - Phase-scoped (discuss/plan/execute/verify) ensure a gsd/phase-NN branch and open a draft PR; report/route run after publish. - Project-level (map-codebase, new-project, new-milestone, complete-milestone) ensure a date-stamped gsd/* branch and open a ready-for-review PR; complete-milestone gains a branch node ahead of archive. - gsd-ship edits + marks ready an existing PR instead of skipping it, and refuses to ship from the base branch. - gsd-quick / gsd-debug ship their work as a PR. All except gsd-debug (diagnose-only, must stay usable without GitHub) declare requires: [github]. Branch logic leaves an existing feature branch untouched and fails fast on the base branch when no phase arg is given. Headers, descriptions, and the coleam00#1696 CHANGELOG entry updated.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
.archon/workflows/experimental/gsd-*.yaml) plus 10 shared role command files (.archon/commands/gsd/gsd-*.md) and a.planning/.gitignore.UX Journey
Before
After
Architecture Diagram
Before
After
Connection inventory:
Label Snapshot
risk: lowsize: Lworkflowsworkflows:experimentalChange Metadata
featureworkflowsLinked Issue
Validation Evidence
okwith zero issues.bun run validate(pre-existing pi-vendor-map staleness halts pipeline);bun run lint(pre-existing OOM);bun run test(no TypeScript changes, YAML/MD only).Security Impact
.planning/and staging area, same as existing workflows)Compatibility / Migration
Human Verification
cli validate workflowswith zero errors and zero warningscli workflow listagents:field (not supported) — models are instructed to read role files insteadapproval:schema withcapture_responseandon_rejectworktree: { enabled: false }on all 11 workflowsrequires: [github]only on gsd-shipinteractive: truecorrectly set on 7 workflows, omitted on 4Side Effects / Blast Radius
gsd-*). They don't affect bundled defaults or existing workflows.cli validate workflowscatches schema issues at load time. Workflow discovery errors are surfaced incli workflow list.Rollback Plan
git revert <commit>— single commit, no schema migrations, no code changes.Unknown workflow: gsd-*) if files are missing. YAML validation errors if schema violations.Risks and Mitigations
worktree.enabled: falseandmutates_checkoutdefaults totrue(serialized). Users are expected to branch before execute, as in GSD's own model.Deviations from Issue #1696
Flat placement:
.archon/workflows/experimental/gsd-*.yamlinstead ofexperimental/gsd/*.yaml. Reason: workflow discovery deliberately caps at depth 1 (workflow-discovery.ts:89). A subdirectory would be silently skipped. The depth cap is intentional and documented — not a bug to fix in this PR.No isolation worktrees: All 11 workflows run with
worktree.enabled: false. The issue's user-flow sketch mentioned verify-work running "in an isolation worktree," but.planning/state and unpushed execute commits would not be present in a fresh worktree. GSD's own model has users branch manually before execute — ship pushes that branch. Same model here.gsd-quick + gsd-debug: Two additional workflows beyond the issue's 9 core commands (per user request in the issue thread).
Promotion criteria: These are experimental — not bundled defaults. Promotion to
.archon/workflows/defaults/is gated on N real-world full cycles, tracked in Port GSD framework to Archon workflows (experimental/gsd/) #1696.Summary by CodeRabbit
Documentation
Chores