Prior 5.5 prompt enumerated rules across 60+ bullets and 312 lines, which
fights GPT-5.5's strength: it follows prose instructions reliably and does
not need rule-by-rule cataloging. Rewrite as flowing paragraphs while
preserving the deep-worker identity and every load-bearing behavior.
Identity preserved:
- Forge god mythology ("Your boulder is code", "forge it until done")
- Direct executor, not orchestrator (research subagents only)
- Senior-colleague tone, end-to-end persistence
Behaviors preserved (compressed to prose):
- Three-attempt failure protocol → 1 paragraph
- Anti-duplication on parallel exploration
- End-to-end usage gate (interactive_bash / playwright / curl / driver)
- Implementation gate: when delegated, execute directly, no draft loop
Net additions distilled from Amp + Codex 5.2 evolution:
- Pragmatism block: smallest correct change, duplication > premature
abstraction, do not over-engineer, do not validate impossible scenarios
- Default-no-tests: add tests only when user asks, fixes a subtle bug,
or protects an important boundary; never to codebases without tests
- WIP-not-legacy: earlier unreleased shapes in the same turn are drafts,
not legacy contracts requiring backward compatibility
- Multi-agent worktree: continue task without reverting unknown changes
- Code-review mode trigger: "review" → findings-first, severity-sorted
- Personality-first opener (Codex 5.2 pattern) for tonal priming
Metrics: 312 → 110 lines (-65%), 60+ bullets → 4 bullets, 21,803 → 15,654
chars (-28%). 26/26 verification checks pass.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Hephaestus is the autonomous deep-worker - everything it receives is a
delegation. Existing 'Manual behavior' bullet vaguely said 'actually run
it' but left the validation surface unspecified, which lets a checked-in
diff plus passing tests masquerade as completion on user-visible work.
Add a dedicated 'End-to-end usage is the gate' subsection in Codex prose
style (no threats/CAPS, contract frames). Surface determines tool:
- TUI / CLI → interactive_bash (tmux), drive it like a real user
- Web / browser / UI → playwright skill, drive a real browser session
- HTTP API / service → curl or integration script against running service
- Library / SDK → minimal driver script
Reinforce in Forbidden stops trailer: when receiving a delegation,
execute directly and validate through the gate; do not loop back with
a draft when the work is yours to do.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
The existing FULL DELEGATION manual-QA rule said 'use it yourself' but
left the choice of tool implicit. Make it explicit and non-optional, so
the agent cannot satisfy the gate by reading the source instead of
running the artifact.
Surface → tool mapping:
- TUI / CLI work → interactive_bash (tmux). Launch in real terminal,
send keystrokes, run happy path, try bad input, hit --help.
- Web / browser / UI work → playwright skill. Drive a real browser,
click elements, fill forms, watch console, screenshot if helpful.
- HTTP API / service work → curl or integration script against the
running service.
- Library / SDK work → minimal driver script that imports + executes.
- Other surfaces → ask how a real user would discover it works, then
do that.
Frame the gate as a contract violation when bypassed: reporting
'implementation complete' without using the matching tool is the same
failure pattern as deleting a failing test for a green build.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Existing <verification> required tests pass + lsp clean + build green, but
that is insufficient for end-to-end delegation. Tests cover known cases;
they do not cover whether the user-visible feature actually works.
Add a NON-NEGOTIABLE rule: when the user hands off end-to-end ("ulw",
"implement and finish", "do the whole thing", "make it work", "ship it"),
verification escalates to:
1. BUILD the actual artifact
2. USE IT YOURSELF as a real user would
3. VERIFY end-to-end behavior matches the spec
4. TASK NOT DONE until usage confirms it works
Reporting "implementation complete" without having USED the artifact is
explicitly framed as a contract violation. Defects discovered during this
QA pass are the agent's to fix in the same turn.
This complements the existing 'lsp_diagnostics catches type errors, not
logic bugs' line by giving full-delegation cases a sharper, named gate.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Root cause: when a file:// prompt URI resolves outside the project root, resolvePromptAppend returns the warning '[WARNING: Path rejected: $URI]' with no indication of WHY the path was rejected. Issue #3554 reports that this is confusing because the docs explicitly advertise support for absolute, home-relative, and cross-project file:// paths, yet the code intentionally restricts file:// prompt resolution to the project boundary (commit 98659783, security hardening).
Fix: extend the warning message so it now includes the resolved project root and an explicit hint that file:// prompts must reside within the project boundary. The security restriction itself is preserved unchanged.
Verification: added a regression test that asserts the rejection warning matches /outside project root/i. Test fails before the fix, passes after. Full resolve-file-uri.test.ts suite: 11 pass / 0 fail. typecheck clean.
Root cause: runBunInstallWithDetails() invoked spawnWithWindowsHide() without passing env, so the child bun install lost https_proxy / http_proxy / NO_PROXY and other parent env vars. @npmcli/agent then received an empty proxy URL and rejected fetch with 'fetch() proxy.url must be a non-empty string', breaking plugin auto-install on networks that require an outbound proxy.
Fix: pass env: process.env to spawnWithWindowsHide so the child bun install inherits the full parent environment, including proxy variables.
Verification: added a regression test that sets https_proxy/http_proxy on process.env and asserts those values are forwarded to the spawn options. Test fails before the fix and passes after. typecheck clean. Other pre-existing path-separator test failures on Windows are unrelated to this change.
Document the canonical ordering mechanism (Array.prototype.toSorted/sort shim in src/shared/agent-sort-shim.ts) and forbid ZWSP, U+2060, U+00AD, ANSI escape, and ASCII space prefixes. Reference sst/opencode#19127 as the upstream fix that will obsolete the shim.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
The sort shim from the previous commit enforces canonical core ordering at runtime, so ZWSP prefixes are no longer needed. Removing them eliminates the Bun.stringWidth vs terminal-width drift that broke the TUI status bar (#3259).
Drop AGENT_LIST_SORT_PREFIXES and getAgentRuntimeName from agent-display-names; switch all call sites to getAgentDisplayName. getAgentListDisplayName stays as a thin alias for external importers.
Keep stripInvisibleAgentCharacters and the ZWSP regex paths so legacy session state and configs from v3.14.0-v3.16.0 still resolve.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
OpenCode 1.4.x ignores the agent `order` field (sst/opencode#19127), so its
`Agent.list()` sorts purely by `agent.name` via Remeda `sortBy` which uses
native string `<`/`>` comparison. Without intervention, the four core agents
fall into alphabetical order (Atlas -> Hephaestus -> Prometheus -> Sisyphus),
which is not the canonical sisyphus -> hephaestus -> prometheus -> atlas order
the project ships.
Prior attempts to bias the sort key with invisible characters (ZWSP,
U+2060 WORD JOINER, U+00AD SOFT HYPHEN, ANSI escape) all caused
`Bun.stringWidth()` vs terminal-width drift, producing visible gaps and
column truncation in the TUI status bar (#3259, #3238).
Solution: a narrowly-scoped shim of `Array.prototype.toSorted` and
`Array.prototype.sort` that activates only when the array contains two or
more agent objects whose `.name` matches a canonical core display name.
The activation predicate guards against mixed-type arrays so unrelated
`.sort()` / `.toSorted()` calls (string arrays, number arrays, mixed
objects) execute native behavior unchanged. Install is idempotent.
Cubic P1 mitigations from PR #3267:
- `isAgentArray` rejects any array with non-object or null elements,
eliminating the throw-on-mixed-array failure mode.
- Strict activation predicate (>= 2 ranked elements) keeps the global
prototype patch from affecting unrelated sort calls.
Remove this shim once OpenCode honors the agent `order` field
(sst/opencode#19127).
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
YeonGyu-Kim