PR #7 ran under the AI-USAGE label and reported real numbers: 28 agent steps,
348s, 2,071,025 input / 17,303 output tokens, and zero cache reads or writes.
The model predicted ~$0.73 on Opus 5 for that tier. The measurement prices it
at $10.79 — the model was ~15x low.
Two wrong assumptions:
- Step count and per-step growth. `full` assumed 12 steps and 1,200 tokens per
tool result; the run did 28 steps averaging ~3,300. Cost is roughly quadratic
in steps, so this compounds. Tier defaults are re-derived from the measured
per-step growth rather than from guesses.
- Caching. The model defaulted to prompt caching on. The headroom/glm-5.2 path
reports 0 read / 0 write, so the stable prefix is paid at full input price on
every step. Budget with caching off until that column is nonzero.
Adds OBSERVED_RUNS as an append-only calibration anchor, an observed-runs
section in the report, and a regression test asserting the model stays within
2.5x of the measurement — so the next drift is caught by the suite rather than
by a surprising invoice.
Corrected blended figures at 350 PRs/month: ~$1,740 Opus 5, ~$1,755 GPT-5.6
Sol, ~$696 Sonnet 5, ~$348 Haiku 4.5, ~$70 GPT-5.6 Luna.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B11e8TZZxJyzHW7jj7KWUN
Skills — the primary now loads conditionally (each one is input tokens), per a
load table in pragent.md:
- attention-tiering: classify every PR trivial/lite/full/oversized BEFORE
reading anything, and cap file reads, linter runs and subagent fan-out per
tier. This is the cost governor; the other skills defer to its budget.
- linter-playbook: per-ecosystem detect-and-run commands scoped to changed
files, the never-install rule, and how to turn a diagnostic into a finding
instead of pasting tool output.
- security-lens: the inline security checklist for when @security isn't worth
delegating, built around a source -> sink test each finding must pass.
- malicious-change: hostile-PR detection — injection aimed at the reviewer,
install/CI-time hooks, obfuscated payloads, dependency confusion, logic
backdoors. Complements the runtime containment added in the previous commit:
that stops the agent being hijacked, this makes it report the attempt.
- comment-craft: how to write problem/fix/suggestion so a maintainer can act in
one read, and what to cut.
pilot/cost_model.py — prices a review against published Claude and OpenAI rates
(fetched 2026-08-18). Prompt sizes are measured from the factory files rather
than guessed; per-tier workloads come from the tiering budgets. The model is
explicit about the thing that actually dominates an agent loop: the whole
conversation is resent every step, so caching moves ~2.3x of the bill.
Blended over a 5/35/55/5 mix with caching on: ~$0.61/PR on Opus 5 or GPT-5.6
Sol, ~$0.24 on Sonnet 5 or Terra, ~$0.12 on Haiku 4.5, ~$0.02 on Luna. At 350
PRs/month that's ~$212 / ~$85 / ~$43 / ~$8.50.
Tests: 101 -> 122.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B11e8TZZxJyzHW7jj7KWUN
The reviewer runs an opencode agent with `bash: "*": allow` over a checkout of
the PR author's branch, and the pod holds a Gitea Write credential. Those two
facts had no wall between them.
Security
- _build_env now allow-lists the subprocess environment instead of inheriting
it, so PRAGENT_BOT_TOKEN and WEBHOOK_SECRET never reach the agent. This was
the live hole: a PR body or an AGENTS.md could ask the agent to `curl` the
token out, and it had both the value and the tool.
- sanitize_workdir deletes author-controlled agent-instruction files from the
checkout before opencode starts (AGENTS.md at any depth, CLAUDE.md,
.cursorrules, a repo opencode.json/.opencode, copilot-instructions.md).
opencode loads nested AGENTS.md as instructions, so a PR could otherwise ship
its own system prompt. They are still reviewed, as data.
- The brief fences PR title/body and diff in --- UNTRUSTED --- markers under a
trust-boundary preamble; the pragent agent, the three lens subagents and the
review-methodology skill now treat injection attempts as a critical finding
to report rather than an instruction to obey.
- .pr-review.json is read from the PR's base branch, not the head sha. Its
`instructions` field is spliced into the reviewer's prompt, so head-ref
reading let any author rewrite the reviewer's rules. Fields are length-capped.
- Untar rejects escaping symlinks, parent traversal, and writes through a
planted symlink (tar-slip).
- The image runs as uid 10001 instead of root.
Robustness
- Bounded review concurrency (PRAGENT_MAX_CONCURRENT_REVIEWS, default 2). Each
review forks an opencode process; a thread per delivery was a fork bomb on a
burst of labels or Gitea retries.
- An in-flight (repo, index, sha) claim closes the check-then-act race in the
sha-marker dedupe, where two deliveries a second apart both read "not yet
reviewed" and both posted.
- Request bodies are capped before being read into memory.
Correctness
- parse_diff_anchors counts a whitespace-stripped blank context line. Skipping
it desynced the new-line counter for the rest of the hunk and silently
misplaced every later inline comment in that file.
- post_inline_review's body-only fallback folds the anchored findings into the
body. It previously posted a summary saying "N inline comment(s) below" with
no comments and no findings — losing them all on the one path that matters.
- fetch_pr_diff's files-endpoint fallback emits real a// b/ prefixes (so
changed_files and the anchor parser work on it) and reports both HTTP statuses
in its error instead of the same one twice.
- The CI workflow template pins PRAGENT_ENGINE=ollama; review_pr defaults to
opencode, which does not exist on a Gitea Actions runner.
Tests: 68 -> 101, covering each of the above.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B11e8TZZxJyzHW7jj7KWUN
- pilot/webhook_server.py: stdlib HTTP receiver. HMAC-verifies X-Gitea-Signature,
gates on pull_request action + AI-REVIEW label, runs review_pr in a background
thread (responds 202 immediately so Gitea's delivery timeout never fires).
Accepts both GitHub-style (labeled/synchronize) and Gitea event-type-style
(label_updated/synchronized) action names.
- pilot/ai_review.py: extract review_pr() core so both the CI run() and the
webhook server share one review path. run() is now an env-driven wrapper.
- pilot/README-webhook.md: architecture, onboarding, one-time per-owner
user-webhook setup, the Gitea 1.26.1 system-webhook bug, the SSRF
ALLOWED_HOST_LIST change, K8s deploy + script-update recipe.
- README.md + design doc: note the webhook service as the preferred delivery
path (partially reverses 'central webhook = non-goal', pilot only).
Gitea 1.26.1 system webhooks broken (POST /admin/hooks -> 201 but never
persists); user-level webhooks (one per repo-owner) are the working fallback.
Gitea SSRF allow-list blocks in-cluster webhook targets by default; required a
scoped [webhook] ALLOWED_HOST_LIST addition + gitea restart.
E2E verified 2026-08-17: pragent-bot reviewed gitea_admin/pragent PR #2 and
masi/portfolio PR #3 via the webhook service (glm-5.2:cloud).
Co-Authored-By: Claude <noreply@anthropic.com>
Ships a working pragent pilot ahead of the framework build (design doc
deferred). Single stdlib-only reviewer script fetched at runtime by a per-repo
Gitea Action; reviews fire only on PRs with the AI-REVIEW label; model is
glm-5.2:cloud via the on-network headroom proxy; fail-open.
- pilot/ai_review.py: fetch PR diff, call model, post review as pragent-bot
- pilot/workflow-template.yml: per-repo Gitea Action gated on AI-REVIEW label
- pilot/README.md: onboarding (bot collaborator + secret + workflow + label)
- tests/pilot/test_ai_review.py: 15 unit tests for pure helpers (no network)
- README/design doc: note pilot is the bootstrap; framework build deferred
Co-Authored-By: Claude <noreply@anthropic.com>
Marcos