OpenAI and the reported partner platforms will implement a shared agent-skills standard that enables practical capability portability across otherwise competing agent ecosystems.

state: expiredheat: lowuncertainty: highconvergesscott: mediumagent-interoperability agent-skills mcpOpenAI

What is this?

Secondary reports describe an open “Agent Skills” format for packaging reusable agent capabilities, with Anthropic publishing a specification and SDK and OpenAI reportedly using a structurally similar directory and metadata scheme in ChatGPT and Codex CLI. The claimed benefit is portability of procedural capabilities across competing agent platforms, complementing MCP’s tool-connectivity role. However, the supplied snippets do not establish a formal agreement between OpenAI and four rivals, and they conflict on whether this is one explicitly adopted standard or independently convergent formats; the cited “Open Plugin Specification” commit is also too truncated here to verify its relationship.

Why it matters to Scott

OpenAI’s reported adoption of a structurally similar skills package would be a consequential convergence with Scott’s Markdown OS, reusable-skills, and vendor-independent portability positions, creating a dated-receipts publishing opportunity. However, the supplied evidence does not yet verify a shared standard or formal cross-vendor agreement, so the stronger interoperability claim remains provisional.

ip:framework.markdown-osip:concept.skills-and-workflowsip:concept.vendor-lock-indev:concept.markdown-native-workflowradar:concept.agent-interoperabilityradar:anthropic-claude-5-context-engineeringradar:copilot-review-skills-mcp-validation

queries asked of Scott's wikis

portable agent skills versus MCP tool connectivity
agent capability packaging and progressive disclosure
cross-platform coding-agent workflow portability
open standards versus vendor lock-in for agent ecosystems
skills as procedural memory for agents
agent plugin and tool interoperability architecture

Measured heat

no measured readings yet — the hourly heat pass fills this in

How the heat travelled

no chain yet — the hourly chain pass fills this in

Evidence (6) — ⭐ canonical anchor

source	object	author	score	comments
🟧 hn	OpenAI and four rivals just agreed on one standard for AI agents	FireBeyond	24	3
🟧 echo.github ⭐	The earliest primary artifact is the repository commit “Add Open Plugin Specification v1.0.0,” described as a “Minimal standard for packagin	John Lindquist	—	—
🟠 reddit	I benchmarked my skill pack against GSD and Superpowers and published the categories where it loses ClaudeAI	jasoncola1	1	1
🟧 hn	Show HN: agent-hop – reverse-engineered session formats to resume in any agent	HetPatel106	1	0
🟧 hn	Trail of Bits Skills Marketplace	wslh	1	0
🟠 reddit	Am I the only one who thinks Anthropic’s Skills system is a little too convenient? ClaudeAI	InfinriDev	0	14

Interpretation history

2026-08-20T15:37:08Z
Repeated rechecks have produced only adjacent implementations and repetitive discussion, with no first-party OpenAI adoption, common vendor specification, or demonstrated cross-platform skill execution. The de facto skills-convergence pattern persists elsewhere, but this specific OpenAI-led shared-standard episode has exhausted its evidence horizon.
2026-08-18T14:48:49Z
The stale recheck adds only trivial engagement to previously assessed adjacent implementations. De facto skills convergence remains plausible, but there is still no first-party vendor adoption, shared specification, or demonstrated cross-platform compatibility supporting the central OpenAI-led hypothesis.
2026-08-16T14:36:02Z
Refreshed comments only repeat known distinctions between declarative skills, progressive disclosure, and deterministic hooks. They add no evidence of OpenAI adoption, a common vendor specification, or working cross-platform skill compatibility, so the central hypothesis remains unverified.
2026-08-16T13:27:30Z
The Anthropic Skills discussion clarifies progressive disclosure and the separation between procedural guidance and deterministic hooks, but it does not materially strengthen the shared-standard hypothesis. Independent reusable-skill implementations support de facto convergence, while OpenAI-led adoption and practical cross-vendor compatibility remain unverified.
2026-08-16T13:22:47Z
evidence attached: reddit.post.1vpvsyz — The concrete description of Anthropic Skills’ progressive-disclosure mechanics is relevant implementation context for judging whether agent skills can become portable across ecosystems.
2026-08-15T20:30:02Z
The recheck adds only minor discussion around the adjacent session-portability adapter, not evidence for shared skill packaging. De facto reusable skills remain plausible, but OpenAI-led standardization, partner adoption, and practical cross-vendor compatibility are still unverified.
2026-08-13T19:41:50Z
The recheck adds only negligible engagement to previously assessed adjacent evidence. De facto reusable-skill packaging remains plausible, but there is still no first-party confirmation of an OpenAI-led standard, partner adoption, or demonstrated cross-vendor skill compatibility.
2026-08-11T19:34:51Z
Trail of Bits adds an independent, credible implementation of skills packaging and distribution, strengthening the case for a de facto reusable-skills ecosystem. It still does not corroborate OpenAI-led standardization, partner adoption, or demonstrated cross-vendor compatibility, so the central hypothesis remains unverified.
2026-08-11T19:23:35Z
evidence attached: hn.story.49263123 — Trail of Bits’ first-party skills repository is an independent artifact showing practical distribution and reuse patterns for portable agent skills.
2026-08-10T20:30:17Z
The stale recheck adds no substantive evidence of vendor adoption, a shared specification, or practical cross-platform skill compatibility. Minor engagement on the adjacent session adapter does not change the case’s meaning, so the underlying portability pattern remains watchable while the OpenAI-led standard claim remains unverified.
2026-08-08T19:33:50Z
Agent Hop adds a second implementation-shaped signal that users want portability across competing coding agents and that adapter-based state transfer is technically possible. It is adjacent to, rather than confirmation of, shared skill packaging, so the OpenAI-led standard and practical cross-vendor compatibility remain unverified.
2026-08-08T19:22:31Z
evidence attached: hn.story.49224884 — A working reverse-engineered adapter layer for resuming sessions across multiple coding agents independently demonstrates demand and partial technical feasibility for agent portability.
2026-08-08T06:27:08Z
The case now has implementation-shaped evidence that Markdown skills can achieve de facto portability across Claude Code and Codex, making the underlying pattern worth watching. It still does not verify an OpenAI-led shared standard, formal vendor adoption, or compatibility among the reported partner platforms.
2026-08-08T06:22:02Z
evidence attached: reddit.post.1vintcc — A cross-Claude-Code-and-Codex skill pack provides contextual evidence that portable, declarative agent skills are emerging across competing ecosystems.
2026-08-07T02:22:11Z
The latest attachment adds no independent confirmation of an OpenAI-led shared standard or cross-platform compatibility; it remains amplification of a third-party packaging specification and an unverified vendor-adoption claim.
2026-08-07T00:23:30Z
grounded: converges/medium — OpenAI’s reported adoption of a structurally similar skills package would be a consequential convergence with Scott’s Markdown OS, reusable-skills, and vendor-i
2026-08-07T00:21:27Z
The origin artifact establishes a concrete third-party packaging specification that combines Agent Skills and MCP, but it does not substantiate the reported OpenAI-led cross-vendor agreement or practical compatibility. The case is now primarily an unverified vendor-adoption claim rather than evidence of achieved interoperability.
2026-08-06T23:31:30Z
grounded: converges/high — If the reported implementations are genuinely compatible, several consequential platform vendors are converging on Scott’s existing position that agent capabili
2026-08-06T23:28:42Z
origin walked (codex/luna, conf 0.96): anchor hn.story.49203443 -> echo.github.a97c7af77a by John Lindquist
2026-08-06T23:26:37Z
case created — The reported cross-company agreement is a bounded interoperability episode distinct from existing cases about individual MCP transports, servers, or host protocols.

Decision trace

08-21 01:37expireRepeated rechecks have produced only adjacent implementations and repetitive discussion, with no first-party OpenAI adoption, common vendor specification, or demonstrated cross-platform skill executio
08-21 01:37alert_silentNo new consequential event occurred; the minor engagement change is unrelated to validating the shared-standard claim and does not merit attention outside a future fresh episode with first-party evide
08-21 01:37alert_routeNo new consequential event occurred; the minor engagement change is unrelated to validating the shared-standard claim and does not merit attention outside a future fresh episode with first-party evide
08-21 01:36drop_targetsquiet through full ladder or over cap 8
08-19 00:48repriceThe stale recheck adds only trivial engagement to previously assessed adjacent implementations. De facto skills convergence remains plausible, but there is still no first-party vendor adoption, shared
08-19 00:48alert_silentNo consequential interoperability event occurred; this can wait for first-party adoption, an inspectable common specification, or verified execution of the same skill package across competing platform
08-19 00:48alert_routeNo consequential interoperability event occurred; this can wait for first-party adoption, an inspectable common specification, or verified execution of the same skill package across competing platform
08-17 00:36repriceRefreshed comments only repeat known distinctions between declarative skills, progressive disclosure, and deterministic hooks. They add no evidence of OpenAI adoption, a common vendor specification, o
08-17 00:36alert_silentThe delta is repetitive implementation discussion rather than a new interoperability event; it can wait for first-party vendor adoption, an inspectable shared specification, or verified execution of t
08-17 00:36alert_routeThe delta is repetitive implementation discussion rather than a new interoperability event; it can wait for first-party vendor adoption, an inspectable shared specification, or verified execution of t
08-17 00:21sensor_dirtycomment_update
08-16 23:27repriceThe Anthropic Skills discussion clarifies progressive disclosure and the separation between procedural guidance and deterministic hooks, but it does not materially strengthen the shared-standard hypot
08-16 23:27alert_silentThe new evidence is implementation context rather than a consequential interoperability event; it can wait for first-party vendor adoption, an inspectable common specification, or verified execution o
08-16 23:27alert_routeThe new evidence is implementation context rather than a consequential interoperability event; it can wait for first-party vendor adoption, an inspectable common specification, or verified execution o
08-16 23:23alert_silentThis low-engagement Reddit critique adds no established implementation, access, or cross-vendor interoperability change. It discusses known tradeoffs between progressively disclosed skills and determi
08-16 23:23alert_routeThis low-engagement Reddit critique adds no established implementation, access, or cross-vendor interoperability change. It discusses known tradeoffs between progressively disclosed skills and determi
08-16 23:22attachThe concrete description of Anthropic Skills’ progressive-disclosure mechanics is relevant implementation context for judging whether agent skills can become portable across ecosystems.
08-16 23:22propose_attachThe concrete description of Anthropic Skills’ progressive-disclosure mechanics is relevant implementation context for judging whether agent skills can become portable across ecosystems.
08-16 06:30repriceThe recheck adds only minor discussion around the adjacent session-portability adapter, not evidence for shared skill packaging. De facto reusable skills remain plausible, but OpenAI-led standardizati
08-16 06:30alert_silentNo consequential new event occurred; minor engagement on adjacent portability tooling can wait for an inspectable common specification, first-party vendor adoption, or verified cross-platform skill ex
08-16 06:30alert_routeNo consequential new event occurred; minor engagement on adjacent portability tooling can wait for an inspectable common specification, first-party vendor adoption, or verified cross-platform skill ex
08-14 05:41repriceThe recheck adds only negligible engagement to previously assessed adjacent evidence. De facto reusable-skill packaging remains plausible, but there is still no first-party confirmation of an OpenAI-l
08-14 05:41alert_silentNo consequential new event occurred; this can wait for an inspectable common specification, first-party vendor adoption, or verified cross-platform execution.
08-14 05:41alert_routeNo consequential new event occurred; this can wait for an inspectable common specification, first-party vendor adoption, or verified cross-platform execution.
08-12 05:34repriceTrail of Bits adds an independent, credible implementation of skills packaging and distribution, strengthening the case for a de facto reusable-skills ecosystem. It still does not corroborate OpenAI-l
08-12 05:34alert_silentThe new repository is useful ecosystem evidence but not a time-sensitive interoperability event; it can wait for the next briefing unless a common specification, first-party vendor adoption, or verifi
08-12 05:34alert_routeThe new repository is useful ecosystem evidence but not a time-sensitive interoperability event; it can wait for the next briefing unless a common specification, first-party vendor adoption, or verifi
08-12 05:25alert_silentThe Trail of Bits repository is a concrete ecosystem-adoption signal, but the supplied evidence establishes only that a skills marketplace exists; it does not show a new shared standard, formal cross-
08-12 05:25surface_candidateThe Trail of Bits repository is a concrete ecosystem-adoption signal, but the supplied evidence establishes only that a skills marketplace exists; it does not show a new shared standard, formal cross-
08-12 05:25alert_routeThe Trail of Bits repository is a concrete ecosystem-adoption signal, but the supplied evidence establishes only that a skills marketplace exists; it does not show a new shared standard, formal cross-
08-12 05:23attachTrail of Bits’ first-party skills repository is an independent artifact showing practical distribution and reuse patterns for portable agent skills.
08-12 05:22propose_attachTrail of Bits’ first-party skills repository is an independent artifact showing practical distribution and reuse patterns for portable agent skills.
08-11 06:30repriceThe stale recheck adds no substantive evidence of vendor adoption, a shared specification, or practical cross-platform skill compatibility. Minor engagement on the adjacent session adapter does not ch
08-11 06:30alert_silentNo consequential new event occurred; the only change is negligible engagement on already-assessed adjacent evidence, so this can wait for first-party adoption, an inspectable common specification, or
08-11 06:30alert_routeNo consequential new event occurred; the only change is negligible engagement on already-assessed adjacent evidence, so this can wait for first-party adoption, an inspectable common specification, or
08-09 05:33repriceAgent Hop adds a second implementation-shaped signal that users want portability across competing coding agents and that adapter-based state transfer is technically possible. It is adjacent to, rather
08-09 05:33alert_silentThe new implementation is an early reverse-engineered session adapter, not evidence that vendors adopted a shared skills standard; its compatibility and durability remain unvalidated, so it can wait f
08-09 05:33alert_routeThe new implementation is an early reverse-engineered session adapter, not evidence that vendors adopted a shared skills standard; its compatibility and durability remain unvalidated, so it can wait f
08-09 05:23alert_silentAgent Hop is an early, low-engagement tool claiming cross-agent session resumption through reverse-engineered adapters. It is practical evidence that some conversation-state portability can be enginee
08-09 05:23alert_routeAgent Hop is an early, low-engagement tool claiming cross-agent session resumption through reverse-engineered adapters. It is practical evidence that some conversation-state portability can be enginee