fix(planning,session-flow): drop arguments: frontmatter from prd and workflow

[kyle-sexton]

Refs: #4001

Summary

The operator declined the arguments: frontmatter field on #4001 and directed the two remaining pilots, planning:prd and session-flow:workflow, off it. This migrates both to the convention in docs/conventions/skill-argument-shape/README.md: the body reads $ARGUMENTS whole.

Fix

• plugins/planning/skills/prd/SKILL.md: remove arguments: [tier]; the action parse reads the first word of $ARGUMENTS as the tier word, and an empty $ARGUMENTS still means ask for the tier.
• plugins/session-flow/skills/workflow/SKILL.md: remove arguments: [mode, modifier]; the first word of $ARGUMENTS is the mode, and after continue the second word is the modifier. Missing or non-auto modifier stays suggest-only.
• argument-hint unchanged in both. No evals, references or tests used $tier, $mode or $modifier.
• planning 0.45.9 -> 0.45.10, session-flow 0.38.25 -> 0.38.26, with CHANGELOG entries.

Verification

• scripts/validate-plugins.sh: all manifests and the catalog validated.
• scripts/check-changelog-parity.sh --check --check-order: pass.
• scripts/check-changed-skills.sh origin/main: 2 skills checked, 0 failed.
• Every *.test.sh under plugins/planning and plugins/session-flow: pass, except plugins/planning/surface/surface.test.sh ui_c.3, which fails identically on origin/main (browser timing, unrelated).
• grep for $tier|$mode|$modifier|arguments: under both plugins: only historical CHANGELOG entries remain.

Related

• docs(conventions): define the skill argument shape — positional action and subject, earned --flag modifiers #4001 (decision and addendum point 2), docs(conventions): define the skill argument shape (#4001) #4768 (convention doc), Adopt named arguments on multi-slot skills; document verified substitution semantics; lint unescaped $N #3543 (added arguments: [tier]).

🤖 Generated with Claude Code

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-29T01:42:17.846153Z	32b811a	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[claude]

Claude finished @kyle-sexton's task in 2m 14s —— View job

---

Code review — /review:code-review

• Gather context (PR diff via gh pr diff 5212, convention doc, repo-wide grep)
• Understand the request (CI code-review lane per custom instructions)
• Review changed files
• Report findings

Scope: 6 files, 2 skills (planning:prd, session-flow:workflow) dropping the arguments: frontmatter field per the skill-argument-shape convention, plus the two plugin version bumps and CHANGELOG entries.

Findings: none. This is a clean, tightly-scoped migration:

• Both SKILL.md bodies now parse $ARGUMENTS whole exactly as the convention prescribes (first word as action/mode, remainder as subject/modifier), and argument-hint is correctly left untouched since it's independent of the arguments: field.
• Confirmed no other SKILL.md in the repo still declares arguments: frontmatter, so this completes the migration rather than leaving a mixed state.
• Grepped both skills' bodies, and session-flow:workflow's evals/evals.json, for stray $tier / $mode / $modifier references — none found, matching the PR's own claim.
• scripts/validate-plugin-contracts.mjs's admitsArguments() check (the thing that gates unescaped $N in argument-taking skills) keys off argument-hint and $ARGUMENTS in the body, not the arguments: field, so removing it doesn't change that gate's behavior for either skill.
• CHANGELOG entries and version bumps follow the existing format in both files; the marketplace catalog doesn't track per-plugin versions, so no update was needed there.
• The PR's described verification commands (validate-plugins.sh, check-changelog-parity.sh, check-changed-skills.sh, the *.test.sh suites) are author-claimed; I did not independently re-run them.

No inline comments filed — nothing meets the bar for a blocking or flagged finding.

[claude]

Claude finished @kyle-sexton's task in 1m 28s —— View job

---

Security review: no findings

• Gather context (PR diff, changed files)
• Check skip-gate conditions (open PR, security-relevant surface, instruction-surface deletion lens)
• Review changed files for security issues
• Post findings or a clean-review statement

Scope reviewed (full diff via git show HEAD, PR's single commit 32b811a):

• plugins/planning/skills/prd/SKILL.md — drops arguments: [tier] frontmatter, rewrites the Action Router paragraph to parse $ARGUMENTS in prose instead.
• plugins/session-flow/skills/workflow/SKILL.md — drops arguments: [mode, modifier] frontmatter, same prose rewrite for mode/modifier parsing.
• plugins/planning/.claude-plugin/plugin.json, plugins/session-flow/.claude-plugin/plugin.json — version bumps only.
• plugins/planning/CHANGELOG.md, plugins/session-flow/CHANGELOG.md — changelog entries only.

Analysis:

• No code execution surface: both SKILL.md bodies use $ARGUMENTS/$tier/$mode/$modifier only as prose the model reads to route its own behavior — confirmed via plugins/planning/skills/prd/SKILL.md#L73 and plugins/session-flow/skills/workflow/SKILL.md#L57. Neither file interpolates these into a shell/run: block, so there's no new injection surface from the reformatting.
• The arguments: frontmatter field being dropped carries no security function to begin with: per docs/conventions/skill-argument-shape/README.md, named arguments are "only positional aliases" and, per the harness docs it cites, the field performs no validation, escaping, or parsing — $ARGUMENTS was already the same raw, unvalidated string either way. Behavior for malicious/adversarial argument content is unchanged before and after this diff.
• Instruction-surface deletion lens: checked the removed arguments: lines against the instruction exception register's Gate 0 classes (irreversible-action, secret-handling, data-integrity, external-publication, legal-compliance, agent-authority). No match — the field was declarative, non-enforcing documentation sugar, not a guardrail, and docs(conventions): define the skill argument shape — positional action and subject, earned --flag modifiers #4001's own addendum declined it specifically because "the field adds no validation." Nothing is left unchecked by its removal.
• Remaining changes (version bumps, changelog prose) have no security surface.

No security issues found in this PR.

[github-actions]

PR body contract — issue linkage

This PR body conforms to the issue-linkage contract. Nothing to do.