Skip to content

rlaope/oh-my-hermes / skills/omh-ai-job-search

omh-ai-job-search

Verified

[omh] Career search to interview: find verified opportunities, tailor evidence-grounded applications and track confirmed outcomes. Use when the user says: ai-job-search, job search, evaluate this job posting, tailor my resume, tailor my cv, prepare for my job interview, track my job applications, draft my cover letter.

references/execution-and-qa.md

View on GitHub
# Host execution, private delivery and verification

## Storage and authority
Resolve an approved private workspace through the active profile outside the product
checkout. Verify exclusion from ancestor Git tracking before writing; do not mutate
another profile or silently change Git/config. Raw CV, profile text, source bodies and
emails stay private, never in fixtures, shared receipts or OMH's metadata-only ledger.
Public/group replies contain safe counts and next actions: no career text, local paths or private references.
If no approved private destination exists, ask before storing/delivering documents.
Reuse available native retrieval, document and citation tools; no new provider, paid
retrieval, external sync, recurring job or outreach is authorized by this method.

## Actually execute the companion
Read toolkit/README.md for schema, limits and evidence types. Use the host terminal,
Python >=3.11 on macOS/Linux, an approved cwd and a finite timeout (default 40s).
Windows or absent no-follow/dir-fd support: unsupported_platform, exit 2, no write.
In the installed skill directory substitute PYTHON, the approved private absolute
/PRIVATE path and JOB_ID from merge output (operator commands, not OMH CLI):

```sh
PYTHON toolkit/scripts/job_state.py validate --input /PRIVATE/jobs.json
PYTHON toolkit/scripts/job_state.py merge --workspace /PRIVATE/tracker --input /PRIVATE/jobs.json
PYTHON toolkit/scripts/job_state.py summary --workspace /PRIVATE/tracker
PYTHON toolkit/scripts/job_state.py transition --workspace /PRIVATE/tracker --job-id JOB_ID --state draft --evidence-type draft_artifact --evidence-ref applications/JOB_ID/v1/resume.md --occurred-at 2026-10-23T03:00:00Z
```

The occurrence timestamp above is illustrative: supply the actual aware timestamp.
After explicit report/evidence of a completed event, invoke transition again with its
valid state, --evidence-type, --evidence-ref and --occurred-at, then summary. Do not
run a submitted example merely to complete this procedure. Synthetic evaluation events
must be conspicuously labeled synthetic and must not imply a real application send.
The helper performs local validation, merge, transitions and counts: no network,
process launch, provider import, CV ingestion, retrieval, drafting or submission.
A host must actually run it, not a prepared card or displayed command alone. It is
inert during discovery/install. No mandatory provider/LaTeX/Bun dependency exists.

Record argv/interpreter, stdout/stderr and actual exit; read back tracker/state.json
and summary. Exit 0 proves a local operation only, not live retrieval or submission.
On exit 2 (argument/schema/type/limit/identity/path), exit 3 (writer conflict), or
exit 4 (I/O/atomic failure), report the error and preserve old state. Never auto-break
locks, delete history, relax bounds or retry indefinitely. Helper safe paths and typed
checks cannot authenticate user reports or evidence documents; host must verify both.
At documented finite limits stop and ask for a separately scoped workspace. An empty
tracker does not prove a successful search. Counts must come from parsed helper JSON.

## Verify requested outcomes, not instruction presence
Read every requested actual file back: profile/preferences, jobs/shortlist, evidence
map/application/review, interview/gaps and tracker/next-action summary as applicable.
Check evidence references, unknown constraints, source failures/freshness, immutable
versions and reviewer identity. Separate technical package tests from output-quality
review. Text delivery tests establish instructions shipped, not observed behavioral
compliance, privacy, injection resistance or useful documents. A live host evaluation
must follow the generated skill with synthetic candidate facts, real source receipts
and observed artifacts. All-live-blocked is a discovery blocker. Keep upstream
same-task comparison unobserved unless actually run; no parity claim follows from
source inspection. No external actions or real submissions are needed for QA.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 196kUpdated Claude CodeCodexVerifiedInstall