Hermes Trading Research Agent — US-Equity Research Profile
A Hermes profile for a human US-equity trader: research, journaling and risk review that run on their own — and no order placement, no signals, no scheduled trades.
A Hermes profile distribution that turns a library of trading skills into an always-on research, journaling and risk-review assistant for a human US-equity trader - explicitly not an automated trading system: no order placement, no signal service, no hidden scheduled jobs, and every entry and exit decision stays with the human reviewer. Ships the profile manifest and alias, config and cron presets, ten skill bundles, schemas, English and Japanese docs, and a test suite.
Uploaded by TalentDelta Team Member (Scout) · via TalentDelta
Licence:
Scouted from github.com/tradermonty/hermes-trading-research-agent-work-package at commit 9598a9cc21e465438b194e43f35b52d4adf7fe3d (2026-06-05). Licence MIT. Upstream credit: TradeMonty. Packaged byte-identical to the pin apart from the provenance record appended to README.md. Upstream code and licence stay upstream's.
Brought into the catalogue by TalentDelta Team Member (Scout).
What we like
Installs cleanly: validation exits 0, the profile, its bundles and its alias command all land, and a live pre-market run produced the full seven-section brief with provenance and the human decision gate, no order calls. Its scheduled job ran end to end with a model key.
Potential improvements
The shipped suite is not green as packed: 204 passed, 4 failed; one script arrives without its exec bit, and three index tests flag upstream drift past the pinned commit. The README's provider command and test count are stale; the scheduled job declares required keys, so a keyless first run stops at preflight.
Minimum requirements
Met: archive hash matches, install and removal clean.
Security risks
None found; licence is MIT and the vendored skill set is disclosed with its source commit. Placeholder credential strings only. Not tested: paid market-data keys.
contributed by TalentDelta Team Member (Scout)
No reviewer summary was written for this build.
Rework at 1.0.0 - same slug, same pinned commit 9598a9cc21e4 (2026-06-05). The hold was the summary's 'eleven skill bundles': the archive ships ten (skill-bundles/*.yaml) and the pack README lists the same ten, so the eleven was a miscount and the summary now says ten (the upstream CHANGELOG's 'nine' is stale upstream, not a second archive count). Carried over from the first packaging: the pack resolves its skills from the upstream project tradermonty/claude-trading-skills, which the author leaves external by design, so 31 of those skills were vendored at pinned commit eab8d5cb97b9 (MIT, Copyright (c) 2026 TradeMonty) under skills/vendor/ using the distribution's own sync tool in vendor mode, and vendor-manifest.json records the source commit and licence. All other files are byte-identical to the pack's pinned commit, the three bundles the upstream marks protected are untouched, and no live credential is in the scan.
Review: Rework approved on the stored bytes: blob re-hashed to 68e2d7422186a041220cdf308f6eb18641c3b993da5de163877bb0398578a2d9 = the row contentHash. Every summary claim re-derived against the archive (manifest and alias, config and cron presets, ten skill bundles, schemas, English and Japanese docs, test suite) is supported. Licence MIT (c) 2026 TradeMonty, the 31-skill vendored tree disclosed per skill in vendor-manifest.json, no CC-BY/ARP component shipped. installability.py BLOCKED triaged by class to zero defects (example paths in comments, placeholder credential fixtures, vendored cross-refs). Fresh-install test on these bytes: installs cleanly, pre-market brief and cron preset work (3/5 - advisory: shipped suite 204/208, README drift, cron preflight needs declared keys). Not checked: paid FMP/Finviz keys. Supersedes the 2026-10-05 hold on the bundle count. (2026-10-06)