Compare commits

..
4 Commits
Author SHA1 Message Date
ffaerber 1aa1f6d7cc Merge pull request 'Add a human-merge label that keeps autopilot from merging' (#126) from ai/autopilot-human-merge into main
ci / lint (push) Successful in 12s
2026-09-16 16:29:46 +02:00
Felix FaerberandClaude Opus 5 cbe3e3cbc0 Add a human-merge label that keeps autopilot from merging
ci / lint (pull_request) Successful in 12s
autopilot skips both human checkpoints: approval before the build and the
merge. Scheduled update issues in homelab want the hands-off build but a human
merge, since merging there deploys. With human-merge also on the issue, @pm is
told to hand back at phase 3 and publish.sh refuses MERGE_PR.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-09-16 17:08:27 +03:00
ffaerber 85be0a3288 Merge pull request 'Let caller-provided skills receive their secrets again' (#125) from ai/caller-skill-secrets into main
ci / lint (push) Successful in 14s
Reviewed-on: #125
2026-09-15 09:46:59 +02:00
Felix FaerberandClaude Opus 5 e29b8141f8 Let caller-provided skills receive their secrets again
ci / lint (pull_request) Successful in 17s
#122 gave the reusable workflow an explicit workflow_call secret contract so
XAI_API_KEY would stop arriving empty. That fixed the model key and quietly
broke every caller-provided skill: SECRETS_JSON is toJSON(secrets), so it now
contains only the secrets this contract declares, and a skill's setup.sh finds
nothing under the names its own repo uses.

homelab's node1-ssh skill is the live casualty. Its setup.sh reads SWARM_HOST,
SWARM_USER and SSH_PRIV_KEY_B64, takes the "not set -- skipping node1 SSH alias"
path, and exits 0; the agent then loads the skill, is told it may run
`ssh node1 ...`, and gets nothing. @senior hit exactly this while diagnosing the
immich crash-loop in homelab #301 and had to fall back to a blind version
revert. The failure is silent by design -- the skip is what lets a fork run
without deploy secrets -- so nothing in the run says the capability is gone.

Declaring the three optional here is the smallest fix that keeps the contract.
It is not a general one: secret names for caller skills are by definition known
only to the caller, and an explicit contract cannot express that, so each new
caller-skill secret has to be added alongside these.

Also fixes two pieces of drift that would spread the same breakage: the header
still told callers to use `secrets: inherit`, and README's copy-me stub still
showed it, so the next repo onboarded would have arrived with no model key.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CYFnHida4AdiNDek2pzgXx
2026-09-15 10:09:19 +03:00
7 changed files with 52 additions and 15 deletions

No files matched your search

+15 -1
View File
@@ -1,6 +1,7 @@
name: agent name: agent
# Reusable AI-agent workflow, shared across repos. A caller repo triggers on issue_comment/issues # Reusable AI-agent workflow, shared across repos. A caller repo triggers on issue_comment/issues
# and invokes this via: uses: gitea/agents/.gitea/workflows/agent.yml@main (secrets: inherit). # and invokes this via: uses: gitea/agents/.gitea/workflows/agent.yml@main (explicit secret map,
# see .gitea/workflows/ai-agent.yml for the stub every caller copies).
# The gate + steps run in the caller's event context (github.event.* / github.repository are the caller's). # The gate + steps run in the caller's event context (github.event.* / github.repository are the caller's).
on: on:
workflow_call: workflow_call:
@@ -29,6 +30,17 @@ on:
required: false required: false
TOKEN_INTERN: TOKEN_INTERN:
required: false required: false
# Secrets for caller-provided skills (.gitea/agent-skills/<name>/setup.sh). Their names are
# known only to the caller repo, but SECRETS_JSON below is toJSON(secrets) and can therefore
# only ever contain what this contract declares — so every such secret has to be listed here,
# optional, or the skill silently no-ops and the agent loses the capability without an error.
# These three are homelab's node1-ssh skill; add further caller-skill secrets alongside them.
SWARM_HOST:
required: false
SWARM_USER:
required: false
SSH_PRIV_KEY_B64:
required: false
# Pinned opencode version — used to install it and to key the CI cache below. # Pinned opencode version — used to install it and to key the CI cache below.
env: env:
@@ -256,6 +268,7 @@ jobs:
HAS_IMAGES: ${{ steps.imgs.outputs.has_images }} HAS_IMAGES: ${{ steps.imgs.outputs.has_images }}
BRANCH: ${{ steps.prep.outputs.branch }} BRANCH: ${{ steps.prep.outputs.branch }}
AUTOPILOT: ${{ steps.prep.outputs.autopilot }} # 'true' when the issue carries the `autopilot` label AUTOPILOT: ${{ steps.prep.outputs.autopilot }} # 'true' when the issue carries the `autopilot` label
AUTOMERGE: ${{ steps.prep.outputs.automerge }} # 'false' when `human-merge` is also set
NUM: ${{ github.event.issue.number }} NUM: ${{ github.event.issue.number }}
TITLE: ${{ github.event.issue.title }} TITLE: ${{ github.event.issue.title }}
IBODY: ${{ github.event.issue.body }} IBODY: ${{ github.event.issue.body }}
@@ -293,6 +306,7 @@ jobs:
NEW: ${{ steps.prep.outputs.new }} NEW: ${{ steps.prep.outputs.new }}
IS_PR: ${{ github.event.issue.pull_request }} # set when this run is on a PR thread IS_PR: ${{ github.event.issue.pull_request }} # set when this run is on a PR thread
AUTOPILOT: ${{ steps.prep.outputs.autopilot }} # 'true' when the origin issue carries `autopilot` AUTOPILOT: ${{ steps.prep.outputs.autopilot }} # 'true' when the origin issue carries `autopilot`
AUTOMERGE: ${{ steps.prep.outputs.automerge }} # 'false' when `human-merge` is also set
ISSNUM: ${{ steps.prep.outputs.issnum }} # origin issue number (resolved from branch on PR threads) ISSNUM: ${{ steps.prep.outputs.issnum }} # origin issue number (resolved from branch on PR threads)
run: bash "$SCRIPTS/publish.sh" run: bash "$SCRIPTS/publish.sh"
+5
View File
@@ -26,3 +26,8 @@ jobs:
TOKEN_QA: ${{ secrets.TOKEN_QA }} TOKEN_QA: ${{ secrets.TOKEN_QA }}
TOKEN_OPS: ${{ secrets.TOKEN_OPS }} TOKEN_OPS: ${{ secrets.TOKEN_OPS }}
TOKEN_INTERN: ${{ secrets.TOKEN_INTERN }} TOKEN_INTERN: ${{ secrets.TOKEN_INTERN }}
# Secrets for this repo's own caller-provided skills (.gitea/agent-skills/). Unset in a repo
# that ships no such skill, which is harmless: the skill's setup.sh no-ops on an empty value.
SWARM_HOST: ${{ secrets.SWARM_HOST }}
SWARM_USER: ${{ secrets.SWARM_USER }}
SSH_PRIV_KEY_B64: ${{ secrets.SSH_PRIV_KEY_B64 }}
+3 -2
View File
@@ -4,7 +4,7 @@
# Required env (provided by the workflow step): # Required env (provided by the workflow step):
# GT TOKEN_PM TOKEN_SENIOR TOKEN_JUNIOR TOKEN_LEAD TOKEN_QA # GT TOKEN_PM TOKEN_SENIOR TOKEN_JUNIOR TOKEN_LEAD TOKEN_QA
# NAME MODE NUM TITLE BRANCH NEW GITHUB_SERVER_URL GITHUB_REPOSITORY # NAME MODE NUM TITLE BRANCH NEW GITHUB_SERVER_URL GITHUB_REPOSITORY
# IS_PR AUTOPILOT ISSNUM (autopilot: @qa label-gated merge/halt + auto-trigger @qa on a fresh PR) # IS_PR AUTOPILOT AUTOMERGE ISSNUM (autopilot: @qa label-gated merge/halt + auto-trigger @qa on a fresh PR)
set +e # publish is best-effort: a grep-no-match / curl non-zero must NOT kill the step set +e # publish is best-effort: a grep-no-match / curl non-zero must NOT kill the step
# Post/PR as the agent's OWN Gitea user when its token is configured; else the built-in bot. # Post/PR as the agent's OWN Gitea user when its token is configured; else the built-in bot.
case "$NAME" in case "$NAME" in
@@ -222,9 +222,10 @@ if [ "$MODE" != "pr" ]; then
fi fi
# --- @pm autopilot merge: @pm is the ONLY agent that merges, and ONLY under the autopilot label --- # --- @pm autopilot merge: @pm is the ONLY agent that merges, and ONLY under the autopilot label ---
# (and never when `human-merge` is also set — AUTOMERGE is false then).
# (@qa never merges — it approves and hands back here.) Merge with the PAT (TTOK), not the built-in # (@qa never merges — it approves and hands back here.) Merge with the PAT (TTOK), not the built-in
# token, so the push to main fires the deploy. TOKEN_PM must carry write:repository. # token, so the push to main fires the deploy. TOKEN_PM must carry write:repository.
if [ "$NAME" = "pm" ] && [ "$AUTOPILOT" = "true" ] && grep -qiE '^[[:space:]]*MERGE_PR[[:space:]]*$' /tmp/agent_out.md; then if [ "$NAME" = "pm" ] && [ "$AUTOMERGE" = "true" ] && grep -qiE '^[[:space:]]*MERGE_PR[[:space:]]*$' /tmp/agent_out.md; then
PRS=$(resolve_prs); PRN=${PRS##* }; CNT=$(echo "$PRS" | wc -w) PRS=$(resolve_prs); PRN=${PRS##* }; CNT=$(echo "$PRS" | wc -w)
if [ -z "$PRN" ]; then if [ -z "$PRN" ]; then
echo "MERGE_PR but no open PR found for issue #$ISSN" echo "MERGE_PR but no open PR found for issue #$ISSN"
+8 -6
View File
@@ -44,7 +44,7 @@ if [ -z "$name" ]; then
# mode=skip so every later step no-ops — a red run for a non-agent comment is just noise. # mode=skip so every later step no-ops — a red run for a non-agent comment is just noise.
echo "no known agent mentioned (word-boundary) — skipping run" echo "no known agent mentioned (word-boundary) — skipping run"
{ echo "name=none"; echo "model=none"; echo "fallback="; echo "vision=false"; echo "mode=skip"; echo "skills=[]"; { echo "name=none"; echo "model=none"; echo "fallback="; echo "vision=false"; echo "mode=skip"; echo "skills=[]";
echo "branch=main"; echo "new=false"; echo "autopilot=false"; echo "issnum=$NUM"; } >> "$GITHUB_OUTPUT" echo "branch=main"; echo "new=false"; echo "autopilot=false"; echo "automerge=false"; echo "issnum=$NUM"; } >> "$GITHUB_OUTPUT"
exit 0 exit 0
fi fi
fi fi
@@ -121,10 +121,12 @@ fi
issnum="$NUM" issnum="$NUM"
case "$IS_PR" in ?*) issnum=$(printf '%s' "$branch_ref" | sed -nE 's,^ai/issue-([0-9]+).*,\1,p');; esac case "$IS_PR" in ?*) issnum=$(printf '%s' "$branch_ref" | sed -nE 's,^ai/issue-([0-9]+).*,\1,p');; esac
[ -z "$issnum" ] && issnum="$NUM" [ -z "$issnum" ] && issnum="$NUM"
autopilot=false # `human-merge` alongside it keeps the hands-off build but leaves the merge to a human.
if curl -sS -H "Authorization: token $GT" "$API/issues/$issnum/labels" 2>/dev/null \ labels=$(curl -sS -H "Authorization: token $GT" "$API/issues/$issnum/labels" 2>/dev/null)
| jq -e 'any(.[]?; .name=="autopilot")' >/dev/null 2>&1; then autopilot=false; automerge=false
if printf '%s' "$labels" | jq -e 'any(.[]?; .name=="autopilot")' >/dev/null 2>&1; then
autopilot=true autopilot=true
printf '%s' "$labels" | jq -e 'any(.[]?; .name=="human-merge")' >/dev/null 2>&1 || automerge=true
fi fi
echo "autopilot (autopilot label on #$issnum)=$autopilot" echo "autopilot (autopilot label on #$issnum)=$autopilot automerge=$automerge"
{ echo "autopilot=$autopilot"; echo "issnum=$issnum"; } >> "$GITHUB_OUTPUT" { echo "autopilot=$autopilot"; echo "automerge=$automerge"; echo "issnum=$issnum"; } >> "$GITHUB_OUTPUT"
+11 -4
View File
@@ -3,7 +3,7 @@
# plain-text reply (/tmp/agent_out.md) plus the raw event stream (/tmp/events.jsonl). # plain-text reply (/tmp/agent_out.md) plus the raw event stream (/tmp/events.jsonl).
# #
# Required env (provided by the workflow step): # Required env (provided by the workflow step):
# XAI_API_KEY SELF_TOKEN NAME MODEL VISION MODE HAS_IMAGES BRANCH AUTOPILOT NUM TITLE # XAI_API_KEY SELF_TOKEN NAME MODEL VISION MODE HAS_IMAGES BRANCH AUTOPILOT AUTOMERGE NUM TITLE
# IBODY CMT # IBODY CMT
# FILES (the opencode -f image flags, from the imgs step output) # FILES (the opencode -f image flags, from the imgs step output)
# AUTOPILOT is 'true' when the issue carries the `autopilot` label (label-gated autopilot mode). # AUTOPILOT is 'true' when the issue carries the `autopilot` label (label-gated autopilot mode).
@@ -74,14 +74,21 @@ if [ "$MODE" = "comment" ]; then
auto-start any dev — the maintainer @mentions an agent on each sub-issue when ready." auto-start any dev — the maintainer @mentions an agent on each sub-issue when ready."
if [ "$AUTOPILOT" = "true" ]; then if [ "$AUTOPILOT" = "true" ]; then
ACTION="$ACTION ACTION="$ACTION
AUTOPILOT MODE IS ACTIVE (this issue carries the 'autopilot' label) — it changes exactly TWO AUTOPILOT MODE IS ACTIVE (this issue carries the 'autopilot' label) — it changes only the
things for you; everything else above is unchanged: phases below; everything else above is unchanged:
- PHASE 1: do NOT ask '@ffaerber ready to build?'. Present your SHORT plan AND end with a - PHASE 1: do NOT ask '@ffaerber ready to build?'. Present your SHORT plan AND end with a
'DELEGATE: @<dev>' line in the SAME reply. Only skip delegating (and ask @ffaerber) if the task 'DELEGATE: @<dev>' line in the SAME reply. Only skip delegating (and ask @ffaerber) if the task
is genuinely ambiguous or unsafe. is genuinely ambiguous or unsafe."
if [ "$AUTOMERGE" = "true" ]; then
ACTION="$ACTION
- PHASE 3: do NOT ask the human to merge. When @qa has approved, end your reply with EXACTLY - PHASE 3: do NOT ask the human to merge. When @qa has approved, end your reply with EXACTLY
'MERGE_PR' — the automation merges the PR and closes this issue. You are the ONLY agent that 'MERGE_PR' — the automation merges the PR and closes this issue. You are the ONLY agent that
merges, and only here." merges, and only here."
else
ACTION="$ACTION
- PHASE 3 is UNCHANGED: this issue also carries 'human-merge', so when @qa has approved, tell
@ffaerber the PR is ready to merge. Do NOT emit 'MERGE_PR'."
fi
fi fi
fi fi
if [ "$NAME" = "qa" ]; then if [ "$NAME" = "qa" ]; then
+2
View File
@@ -29,6 +29,8 @@ An issue labeled **`autopilot`** runs without the usual human checkpoints:
- **Kill switch:** remove the `autopilot` label at any time. The label is re-read fresh at the - **Kill switch:** remove the `autopilot` label at any time. The label is re-read fresh at the
start of every run, so the next agent turn reverts to normal human-approval behavior. `@qa` also start of every run, so the next agent turn reverts to normal human-approval behavior. `@qa` also
removes the label itself whenever it halts on a bug or a failed merge. removes the label itself whenever it halts on a bug or a failed merge.
- **`human-merge` label** (together with `autopilot`): everything above except the merge — after
`@qa` approves, `@pm` tells the creator the PR is ready and a human merges.
No label (the default) = today's behavior, unchanged. No label (the default) = today's behavior, unchanged.
## Branches & pull requests ## Branches & pull requests
+8 -2
View File
@@ -31,7 +31,8 @@ deploy-host SSH skill, live in the consuming repo under `.gitea/agent-skills/`
4. `@qa` reviews **on the PR** — either recommendations + `BOUNCE: @dev` (dev fixes → `@qa` 4. `@qa` reviews **on the PR** — either recommendations + `BOUNCE: @dev` (dev fixes → `@qa`
re-verifies, direct loop, max 3 rounds) or `APPROVE`. re-verifies, direct loop, max 3 rounds) or `APPROVE`.
5. On approval `@qa` posts the verdict **on the issue** → `@pm` tells the creator *"ready to merge"* 5. On approval `@qa` posts the verdict **on the issue** → `@pm` tells the creator *"ready to merge"*
and a **human merges** — or, with the `autopilot` label, `@pm` merges and closes the issue itself. and a **human merges** — or, with the `autopilot` label, `@pm` merges and closes the issue itself
(unless the issue also carries `human-merge`, which keeps the merge with a human).
`@pm` is the only agent that ever merges, and only under the `autopilot` label (its kill switch: `@pm` is the only agent that ever merges, and only under the `autopilot` label (its kill switch:
remove the label mid-flight and the next step reverts to human control). remove the label mid-flight and the next step reverts to human control).
@@ -87,7 +88,12 @@ on:
jobs: jobs:
agent: agent:
uses: gitea/agents/.gitea/workflows/agent.yml@main uses: gitea/agents/.gitea/workflows/agent.yml@main
secrets: inherit # Explicit map, not `secrets: inherit`: inherit alone has left XAI_API_KEY empty on some Gitea
# cross-owner calls. Copy the full list from .gitea/workflows/ai-agent.yml, which is the stub
# this file abbreviates -- a secret missing from the map does not reach the workflow at all.
secrets:
GITEA_TOKEN: ${{ secrets.GITEA_TOKEN }}
# ... and the OLLAMA_*, XAI_API_KEY, TOKEN_* and caller-skill entries, verbatim.
``` ```
That's the whole per-repo footprint, and it's the minimum a caller can be: the `on:` triggers must That's the whole per-repo footprint, and it's the minimum a caller can be: the `on:` triggers must