Skip to content

docs(adr): 0121 steering on by default - #9032

Open
waynesun09 wants to merge 1 commit into
mainfrom
steer-adr-0121
Open

waynesun09 wants to merge 1 commit into
mainfrom
steer-adr-0121

Conversation

@waynesun09

@waynesun09 waynesun09 commented Nov 2, 2026 •

Copy link
Copy Markdown
Member

Reading order: start with #9030 (ADR 0117), then #9028 (0118) → #9029 (0119) → #9031 (0120) → #9032 (0121). PR numbers do not follow ADR numbers, because the ADRs were renumbered into dependency order; this PR is ADR 0121.

Summary

ADR 0121 of five that rebuild the steering design (#6957) on the entity ADRs now on main (0133, 0141, 0144, 0147–0149, 0151). Docs only.

Decision: Steering is on by default, with a per-harness and a per-repository opt-out. An explicit enabled: false at any layer wins; tuning fields never turn it back on. Also defines the max_steers limit (default 3, range 1–9999).

Steering hands a work-item update to the agent that is still running, so it can adjust before finishing on stale state. After ADR 0144 the next sandbox cycle is the correctness backstop; steering is a fast path inside one cycle.

The series

ADR Title PR
0117 Steer a running cycle with admitted entity updates #9030
0118 Steer runtime interface #9028
0119 Signal GitHub Actions events to a running cycle #9029
0120 Record absorbed steers as agent-run receipts #9031
0121 Steering on by default this PR
  • Merge order: 0117 → 0118 → 0119 → 0120 → 0121. Numbers follow dependency: each ADR cites only lower-numbered ones. Each PR passes the docs checks on its own; links to siblings that are not on main yet are plain text ("ADR 0117 (proposed)") and can become links once they merge.
  • Accepted ADRs touched: Status note on ADR 0106: the default-on pointer now links ADR 0121 instead of PR feat(harness)!: steer by default, opt out with steer.enabled: false #7461.

Implementation dependencies

No code here. Implementation lands with or after:

  • ADR 0151 human_min_role admission;
  • the ADR 0141 filter pipeline and host-side extension path;
  • the ADR 0144 bound-run lifecycle;
  • entity-context-snapshot v2 for the receipt fields (ADR 0120).

The runtime interface (0118) and the run-record signal (0119) can be built first; they have no caller until the rest exists.

Lineage

These replace drafts in the earlier steering stack, which predate the entity ADRs.

Never cancelling the in-flight run (#7007) is already ADR 0106's decision, so it needs no new ADR.

Evidence

Current-behaviour claims in ADR 0121, with the code that implements the layers the switches compose across:

Testing

  • pre-commit run --from-ref origin/main --to-ref HEAD: all hooks pass (ADR status, number and front-matter linters, markdown links, secrets).
  • make lint-md-links: 0 errors.

@qodo-code-review

Copy link
Copy Markdown

PR Summary by Qodo

Document steering as on by default with layered opt-outs

📝 Documentation 🕐 10-20 Minutes

Grey Divider

AI Description

• Accept ADR 0121, making steering the default for eligible harnesses once its dependencies are
 ready.
• Define per-harness and per-repository opt-outs; an explicit enabled: false always wins.
• Link ADR 0106 to the new decision without changing its run-serialization policy.
Diagram

graph TD
  U["Work-item update"] --> E{"Steering supported?"} -->|Yes| C["Layered steer settings"] --> D{"Any enabled false?"} -->|No| A["Running agent"] -->|If needed| N["Next cycle backstop"]
  E -->|No| N
  D -->|Yes| N
Loading
High-Level Assessment

The following are alternative approaches to this PR:

1. Require per-harness opt-in
  • ➕ Avoids delivering steers to agents that cannot act on them.
  • ➖ Leaves silent harnesses exposed to stale in-flight output until individually configured.

Recommendation: Keep the documented default-on decision, gated on runtime readiness, with explicit opt-outs for incompatible harnesses. Per-harness opt-in is safer at rollout but would withhold the fast path from every harness that has not been edited.

Files changed (2) +107 / -2

Documentation (2) +107 / -2
0106-serialize-agent-runs-and-coalesce-subsequent-events.mdPoint the steering status note to ADR 0121 +1/-2

Point the steering status note to ADR 0121

• Replaces a pull-request link with a link to the accepted default-on ADR. ADR 0106’s preserve-and-coalesce decision remains unchanged.

docs/ADRs/0106-serialize-agent-runs-and-coalesce-subsequent-events.md

0121-steering-on-by-default.mdRecord the default-on steering decision and opt-out rules +106/-0

Record the default-on steering decision and opt-out rules

• Accepts steering by default for eligible harnesses, with per-harness and per-repository opt-outs where any explicit 'enabled: false' wins. Documents readiness and forge limits, reasons to opt out, and the expected operational consequences.

docs/ADRs/0121-steering-on-by-default.md

@qodo-code-review

Copy link
Copy Markdown

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📎 Requirement gaps (0)

Grey Divider

Great, no issues found!

Qodo reviewed your code and found no material issues that require review

Grey Divider

Tip of the day
💡 Did you know, you can copy the agent prompt from any finding and feed it to your IDE agent

More tips ↗ | Customize Qodo ↗ | Qodo docs ↗

Grey Divider

Qodo Logo

@fullsend-ai-review

fullsend-ai-review Bot commented Nov 2, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 7:00 PM UTC · Completed 7:11 PM UTC

Commit: 5f6e852 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $1.78

@github-actions

github-actions Bot commented Nov 2, 2026 •

Copy link
Copy Markdown

Site preview

Preview: https://3788788c-site.fullsend-ai.workers.dev

Commit: 47d95956f9094ed0b9a044e78582c9dc51acd20c

@fullsend-ai-review fullsend-ai-review Bot added the risk/moderate PR risk: moderate label Nov 2, 2026
@fullsend-ai-review

fullsend-ai-review Bot commented Nov 2, 2026 •

Copy link
Copy Markdown

<!-- fullsend:risk-assessment -->
Risk Assessment: moderate (2/5)

Details

Documentation-only ADR change with no code or protected paths, elevated to moderate risk by medium blast radius, multi-author contention and change coupling on the modified ADR, and an aging linked issue.

Previous run

<!-- fullsend:risk-assessment -->
Risk Assessment: moderate (2/5)

Details

Documentation-only ADR change with no code or protected paths, elevated to moderate risk by medium blast radius, multi-author contention and change coupling on the modified ADR, and an aging linked issue.

Previous run (2)

<!-- fullsend:risk-assessment -->
Risk Assessment: moderate (2/5)

Details

Documentation-only ADR change with no code or protected paths, elevated to moderate risk by medium blast radius, multi-author contention and change coupling on the modified ADR, and an aging linked issue.

Previous run (3)

<!-- fullsend:risk-assessment -->
Risk Assessment: moderate (2/5)

Details

Documentation-only ADR change with no code or protected paths, elevated to moderate risk by medium blast radius, multi-author contention and change coupling on the modified ADR, and an aging linked issue.

Previous run (4)

<!-- fullsend:risk-assessment -->
Risk Assessment: moderate (2/5)

Details

Documentation-only ADR change with no code or protected paths, elevated to moderate risk by medium blast radius, multi-author contention and change coupling on the modified ADR, and an aging linked issue.

Previous run (5)

<!-- fullsend:risk-assessment -->
Risk Assessment: moderate (2/5)

Details

Documentation-only ADR change with no code or protected paths, elevated to moderate risk by medium blast radius, multi-author contention and change coupling on the modified ADR, and an aging linked issue.

Previous run (6)

<!-- fullsend:risk-assessment -->
Risk Assessment: moderate (2/5)

Details

Documentation-only ADR change with no code or protected paths, elevated to moderate risk by medium blast radius, multi-author contention and change coupling on the modified ADR, and an aging linked issue.

Previous run (7)

<!-- fullsend:risk-assessment -->
Risk Assessment: moderate (2/5)

Details

Documentation-only ADR change with no code or protected paths, elevated to moderate risk by medium blast radius, multi-author contention and change coupling on the modified ADR, and an aging linked issue.

@fullsend-ai-review

fullsend-ai-review Bot commented Nov 2, 2026 •

Copy link
Copy Markdown

Looks good to me

Previous run

<!-- fullsend:review-findings-v2:eyJ2ZXJzaW9uIjoyLCJhY3Rpb24iOiJhcHByb3ZlIiwiZmluZGluZ3MiOlt7InNldmVyaXR5IjoibWVkaXVtIiwiY2F0ZWdvcnkiOiJsb2dpYy1lcnJvciIsImZpbGUiOiJkb2NzL0FEUnMvMDEyMS1zdGVlcmluZy1vbi1ieS1kZWZhdWx0Lm1kIiwiaWQiOiJmX2Y0NjQ3MTI3YWE5MGYzZDQiLCJsaW5lIjo0MywiYWN0aW9uYWJsZSI6dHJ1ZSwiZGltZW5zaW9uIjoiY29ycmVjdG5lc3MifSx7InNldmVyaXR5IjoibWVkaXVtIiwiY2F0ZWdvcnkiOiJpbmNvbXBsZXRlLWRvYyIsImZpbGUiOiJkb2NzL2FyY2hpdGVjdHVyZS5tZCIsImlkIjoiZl8yNDk4YzdhMDE3MzY1ZjgwIiwibGluZSI6NDA1LCJhY3Rpb25hYmxlIjp0cnVlLCJkaW1lbnNpb24iOiJkb2NzLWN1cnJlbmN5In0seyJzZXZlcml0eSI6ImxvdyIsImNhdGVnb3J5IjoiZG9jLXN0eWxlIiwiZmlsZSI6ImRvY3MvQURScy8wMTIxLXN0ZWVyaW5nLW9uLWJ5LWRlZmF1bHQubWQiLCJpZCI6ImZfMzMxZmI2ZDI5ZDVkYThhMSIsImxpbmUiOjI3LCJhY3Rpb25hYmxlIjp0cnVlLCJkaW1lbnNpb24iOiJzdHlsZS1jb252ZW50aW9ucyJ9LHsic2V2ZXJpdHkiOiJtZWRpdW0iLCJjYXRlZ29yeSI6ImF1dGgtYnlwYXNzIiwiZmlsZSI6ImRvY3MvQURScy8wMTIxLXN0ZWVyaW5nLW9uLWJ5LWRlZmF1bHQubWQiLCJpZCI6ImZfZTQ1YzBjZDZmNTU3Mjk3YSIsImxpbmUiOjk3LCJhY3Rpb25hYmxlIjp0cnVlLCJkaW1lbnNpb24iOiJzZWN1cml0eSJ9XSwiZGlzcG9zaXRpb25zIjpbeyJpZCI6ImZfZjQ2NDcxMjdhYTkwZjNkNCIsInN0YXR1cyI6InJlc29sdmVkX2J5X2NoYW5nZSJ9LHsiaWQiOiJmXzI0OThjN2EwMTczNjVmODAiLCJzdGF0dXMiOiJyZXNvbHZlZF9ieV9jaGFuZ2UifSx7ImlkIjoiZl8zMzFmYjZkMjlkNWRhOGExIiwic3RhdHVzIjoicmVzb2x2ZWRfYnlfY2hhbmdlIn0seyJpZCI6ImZfZTQ1YzBjZDZmNTU3Mjk3YSIsInN0YXR1cyI6InJlc29sdmVkX2J5X2NoYW5nZSJ9XX0= -->

Looks good to me

Previous run (2)

<!-- fullsend:review-findings-v2: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 -->

Looks good to me

Previous run (3)

<!-- fullsend:review-findings-v2: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 -->

Looks good to me

Previous run (4)

<!-- fullsend:review-findings-v2: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 -->

Looks good to me

Previous run (5)

<!-- fullsend:review-findings-v2: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 -->

Looks good to me

Earlier findings

  • docs/ADRs/0121-steering-on-by-default.md:98 — auth-bypass, resolved: the opt-out paragraph no longer tells harnesses to opt out so a pre-script gate can see every update, and now states every steer passes the pre-script skip gate before delivery (lines 98-100), with the default gated on that pre-script gate being in place (line 83).
Previous run (6)

<!-- fullsend:review-findings-v2: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 -->

Review

Findings

Medium

  • [auth-bypass] docs/ADRs/0121-steering-on-by-default.md:97 — Steering is on by default, and the ADR handles harnesses with pre-script gates only by asking them to opt out (lines 97-99: "when its pre-script gate must see every update before the model does"). In the base tree, ADR 0144 (lines 178-183) runs the harness pre-script once per cycle, after snapshot assembly, so a steer delivered mid-cycle never passes through that pre-script. With the default on, an unchanged harness whose pre-script filters or skips updates (ADR 0072 skip protocol) will start receiving mid-cycle updates its pre-script never checked, unless a maintainer finds the harness and sets steer: {enabled: false}. The "Not before readiness" clause points to ADR 0119, which is not available to review, so it cannot be confirmed whether it adds a per-update check for pre-script-gated harnesses.
    Remediation: Make the default safe for harnesses that have a pre_script: either treat a harness with a pre_script as not eligible for default-on steering unless it explicitly enables steering, or state that each steer must pass the same pre-script/skip gate as a fresh cycle before delivery, with rejected steers falling back to the ADR 0106 preserve-and-coalesce follow-up. List this as a readiness prerequisite in the ADR.

Earlier findings

  • docs/ADRs/0121-steering-on-by-default.md:43, logic-error — resolved: the ADR text was rewritten in the new head and the current Context/Decision text is internally consistent with issue Steer the in-flight agent run on work-item updates instead of cancelling it #6957.
  • docs/architecture.md:405, incomplete-doc — resolved: docs/architecture.md lines 419-424 now carry the "Steering default (target state)" entry linking ADR 0121.
  • docs/ADRs/0121-steering-on-by-default.md:27, doc-style — resolved: prose lines in the head file are wrapped within 80 columns (only the exempt front-matter title and table row are longer).

Next steps:

  • /fs-fix — agent addresses review findings automatically
  • /fs-fix <your instruction> — agent fixes with your specific guidance
  • Push commits directly — review re-runs automatically on push
  • /fs-fix-stop — disable automatic fix runs for this PR
Previous run (7)

<!-- fullsend:review-findings-v2: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 -->

Review

Findings

Medium

  • [logic-error] docs/ADRs/0121-steering-on-by-default.md:43 — The Context states that issue Steer the in-flight agent run on work-item updates instead of cancelling it #6957 "describes harness.steer as off by default" and that ADR 0106 records this criterion as replaced. The current body of Steer the in-flight agent run on work-item updates instead of cancelling it #6957 says the opposite in its validation criteria: "Steering is on by default. A harness that sets steer: {enabled: false} keeps the ADR 0106 path." The new ADR therefore misstates its origin issue in a permanent record. The ADR 0106 status note that this PR re-links carries the same premise.
    Remediation: Reword the Context (and, if appropriate, the ADR 0106 status note) so the ADR formalizes the issue's default-on validation criterion and opt-out composition, or explicitly attributes the off-by-default wording to an earlier version of the issue rather than its current body.
  • [incomplete-doc] docs/architecture.md:405 — docs/contributing/adrs.md requires that when a new ADR is set to Accepted, docs/architecture.md and related problem docs are updated in the same PR. This PR changes only the two ADR files. The "Agent Dispatch and Coordination Layer" Decided list (lines ~405-414) records ADR 0106/0144 serialization and coalescing but does not mention default-on steering with per-harness and per-repository opt-out.
    Remediation: Add an entry under "Agent Dispatch and Coordination Layer" in docs/architecture.md recording the default-on steering decision and the steer: {enabled: false} harness and repository opt-outs, linking ADR 0121. Add related problem-doc backlinks if a doc substantively discusses steering.

Low

  • [doc-style] docs/ADRs/0121-steering-on-by-default.md:27 — Lines are wrapped well short of the ~80-column guideline in docs/contributing/documentation.md. The companion-decision bullets in Status (lines 27-32) break after 22-32 characters even though each bullet fits on one line, and Context line 36 breaks after 27 characters.
    Remediation: Unwrap each companion-decision bullet (lines 27-32) onto a single line and rewrap Context lines 36-39 to fill close to 80 columns.

Labels: ADR documentation change


Next steps:

  • /fs-fix — agent addresses review findings automatically
  • /fs-fix <your instruction> — agent fixes with your specific guidance
  • Push commits directly — review re-runs automatically on push
  • /fs-fix-stop — disable automatic fix runs for this PR

fullsend-ai-review[bot]

This comment was marked as outdated.

@fullsend-ai-review fullsend-ai-review Bot added type/adr component/docs User-facing documentation labels Nov 2, 2026
@codecov

codecov Bot commented Nov 2, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

@waynesun09

Copy link
Copy Markdown
Member Author

/fs-fix-stop

@github-actions github-actions Bot added the fullsend-no-fix Skip bot-triggered fix agent runs label Nov 2, 2026
@github-actions

github-actions Bot commented Nov 2, 2026

Copy link
Copy Markdown

Fix agent disabled for this PR. Remove the fullsend-no-fix label or use /fs-fix to re-engage.

fullsend-ai-review[bot]

This comment was marked as outdated.

@fullsend-ai-review

fullsend-ai-review Bot commented Nov 3, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 2:56 AM UTC · Completed 3:04 AM UTC

Commit: 29d0ca8 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $1.51

@fullsend-ai-review fullsend-ai-review Bot removed the ready-for-merge All reviewers approved — ready to merge label Nov 3, 2026
@fullsend-ai-review
fullsend-ai-review Bot dismissed their stale review November 3, 2026 03:04

Superseded by updated review

fullsend-ai-review[bot]

This comment was marked as outdated.

@fullsend-ai-review

fullsend-ai-review Bot commented Nov 3, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 3:17 AM UTC · Completed 3:24 AM UTC

Commit: e030918 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $1.26

@fullsend-ai-review fullsend-ai-review Bot removed the ready-for-merge All reviewers approved — ready to merge label Nov 3, 2026
fullsend-ai-review[bot]

This comment was marked as outdated.

@fullsend-ai-review
fullsend-ai-review Bot dismissed their stale review November 3, 2026 03:24

Superseded by updated review

@fullsend-ai-review fullsend-ai-review Bot added the ready-for-merge All reviewers approved — ready to merge label Nov 3, 2026
Turn steering on by default, with a per-harness and a per-repository opt-out.

Assisted-by: Claude (fix), Claude (review)
Signed-off-by: Wayne Sun <gsun@redhat.com>
@fullsend-ai-review

fullsend-ai-review Bot commented Nov 3, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 3:35 AM UTC · Completed 3:42 AM UTC

Commit: 47d9595 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5-5 · Effort: high · Cost: $1.38

@fullsend-ai-review fullsend-ai-review Bot removed the ready-for-merge All reviewers approved — ready to merge label Nov 3, 2026
@fullsend-ai-review
fullsend-ai-review Bot dismissed their stale review November 3, 2026 03:42

Superseded by updated review

@fullsend-ai-review fullsend-ai-review Bot added the ready-for-merge All reviewers approved — ready to merge label Nov 3, 2026

This branch was successfully deployed

1 active and 1 inactive deployments
site-preview — 47d95956 Deployed Nov 3, 2026 by github-actions[bot]
dev — 47d95956 Deployed Nov 3, 2026 by waynesun09 via playback #18699
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

component/docs User-facing documentation fullsend-no-fix Skip bot-triggered fix agent runs ready-for-merge All reviewers approved — ready to merge risk/moderate PR risk: moderate type/adr

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant